AI Solutions for Industries

Advanced quality inspection and data insights
AI for Manufacturing, Automotive, Paper, Chemical, and Energy sectors.
AI self-checkout stations for Restaurants and object recognition Retail businesses.
AI Solutions
for Industries

for Manufacturing for Automotive for Paper Industry Data Insights
Cognitive Services
Advanced quality inspection and data insights.
Cognitive Services for Restaurants
Self-checkout and object recognition.
CFD Suite
AI-accelerated Computational Fluid Dynamics.
Predictive Maintenance
Featured Products

Custom AI Development
AI Services
AI Workshop Edge AI Cognitive Automation HPC
Incubation
Intel Expertise Alveo Expertise
NVIDIA Expertise
brainello Ewa Guard
Document Processing Forestry & Agriculture
+48 508 091 885
+48 505 322 282
welcome@byteLAKE.com

What is AI?
➢ Empowering Computers to Learn
Without Explicit Programming
• AI is a field of study that grants computers the ability to learn autonomously,
without the need for explicit programming.
• It functions based on a trained model derived from a dataset of input observations,
typically historical data, and produces data-driven predictions or decisions
as outputs.
• AI Utilizes Senses:
• Visual Perception: Cameras enable AI to perceive and analyze visual data,
such as inspecting quality, counting objects, monitoring processes,
and detecting anomalies.
• Auditory Perception: Microphones allow AI to hear and interpret sound
data, facilitating tasks like inspecting bearings, identifying issues,
and measuring noise levels.
• Other Data: IoT sensors provide supplementary information on factors like
pressure, temperature, humidity, etc. AI can leverage this data, historic data,
input from experts and other data (i.e. weather forecasts, online data, etc.)
to predict trends, anticipate failures, and assist in preventing downtimes.
(*) – Explanation inspired by publications by Arthur Samuel in 1959
(*)

AI Solution
We need
AI!
“I read an article in
Forbes and if we don’t
have an AI strategy,
we’re going to fall
behind the
competition!”
Data Sources
Big Data
AI Solution
AI Training
AI use cases
Visual inspection
(products,
processes)
for better quality.
Automate complex
tasks for better
performance.
Analyze data from
various sources
for better decisions,
optimal results,
improved quality
and reliability.
Successful AI strategy
AI for Industry 4.0 – blog post series
www.bytelake.com/en/CognitiveServices-toc

AI Strategy
Workshop
Proof of Concept
Development
Problem
Identification
and Solution
Exploration
Our journey in developing AI
solutions commences by
meticulously examining our
clients' unique challenges,
listening to their problem
statements, and
brainstorming initial ideas.
We work closely with our
clients to identify areas
where AI can deliver value,
generating a comprehensive
list of potential AI solutions.
To ensure a clear path forward,
we conduct AI strategy
workshops, offering our clients
a deeper insight into the
transformative capabilities of AI
solutions. During this phase, we
collaboratively chart out
deployment plans, meticulously
assess the available data
resources, and devise strategic
activities aimed at guaranteeing
the seamless integration of AI
into their business operations.
Tailored AI Solution
Deployment
At byteLAKE, we understand
the importance of a gradual
approach. Our AI solutions
are developed in a stepwise
manner, beginning with a
limited functionality proof of
concept. This approach
allows us to validate our
clients' objectives, fine-tune
our algorithms, and ensure
that the solution aligns
perfectly with their needs
before proceeding to full-
scale development.
Upon successful completion
of the proof of concept phase,
we embark on the final leg of
deployment. Here, our AI
product, meticulously trained,
calibrated, and customized to
precisely fit our client's unique
requirements, is primed for
production deployment. This
phase underscores byteLAKE's
commitment to accelerating
time to market, ensuring our
clients swiftly realize the
transformative benefits of AI
within their industries.
Streamlined AI Solution Development Process
byteLAKE’s expertise in AI, byteLAKE’s suite of AI products

byteLAKE’s AI Solutions for Industries
Optimal Deployment with byteLAKE’s AI Products
Data
Management
Data Collection byteLAKE’s AI Product DevOps
AI Consultant
SW Engineer &
AI Consultant
Data Engineer Data Scientist AI Engineer
Data Sources Data Storage Hardware Acceleration Edge Device
Find
Solution
Awareness, Demo,
PoC
Data
Collection
Extracting,
storing, retrieving
Data
Cleaning
Cleaning,
formatting,
labeling
AI Model
Training
Calibration
AI Solution
Deployment
and
Integration
Advanced
quality inspection
& data insights

Who are we?
byteLAKE’s
Research & AI Products

➢ LinkedIn.com/company/byteLAKE
➢ X.com/byteLAKEGlobal
➢ FB.com/byteLAKE/
➢ byteLAKE.com/en/YouTube
➢ Blog
Partners & Clients
“AI already plays a very important role in our daily lives. […]
The application of the Intel® Distribution of OpenVINO™ toolkit
in byteLAKE’s Cognitive Services shows that AI works efficiently
as an actual tool for optimizing company operations. Moreover,
such a combination reduces the barrier of necessary upgrades to IT
infrastructure [...],” said Krzysztof Jonak,
EMEA Territory Sales Director, Intel.
“We’re also working with a number of partners on AI initiatives
that will provide real world solutions for customers. […]
Our collaboration with partners such as Intel, NVIDIA, Mark III
systems, and byteLAKE greatly expands the resources and
expertise we’re able to provide“, said Dr. Bhushan Desam,
Lenovo’s AI Global Business Leader, HPC and AI Business.

Our Team
Marcin Rojek
Co-founder
Enthusiast for merging
research and the latest
scientific achievements with
concrete business needs.
The conversion of data into
valuable insights is an area
where Marcin and his team
constantly seek inspiration
for further developing
the company's products.
Krzysztof Rojek
CTO, DSc, PhD
Bridging the gap between
byteLAKE’s business and the
dynamic domains of research
and academia. Impassioned
advocate and catalyst for ideas
that incubate in the research
sphere and later find their
place in tangible, real-world
business applications.
Our Amazing
Engineering Team
Located in Poland.
Mariusz Kolanko
Co-founder
Responsible for creating
and implementing industrial
solutions for major global
firms. Background in
multinational corporations.
An enthusiast for artificial
intelligence and a proponent
of harnessing its potential
in the business landscape.
John Sedgwick
Customer Success, USA
AI is the future, and we are
starting to see and understand
some of the possibilities now.
AI will grow exponentially
across multiple sectors and
throughout the economy as
technology allows us to deploy
smart tactics and solve difficult
problems. Providing strategy
and tactics to find value and
deliver products to meet
customer and segment needs. +48 508 091 885
+48 505 322 282

byteLAKE among top AI companies in Poland!
"It contains information on practically all meaningful
companies operating in Poland which offer services or
products in the field of modern technologies. We believe this map
will be necessary to help both domestic and international
investors looking for interesting projects in Poland.",
Aleksander Kutela, President of Digital Poland Foundation

Our Research Studies
GO PUBLIC!
More at: byteLAKE.com/en/research

AI
Deployment
Plan
Case
Studies
Science +
business +
industry know-
how
& Partners

Cognitive Services
for Manufacturing for Automotive for Paper Industry Data Insights
Predictive Maintenance
• Image analytics for precise visual inspection
• Sound analytics enabling proactive maintenance
• Wet line analytics
• Seamlessly convert data into actionable insights,
enabling advanced predictive maintenance and risk detection

Cognitive Services
AI turning data into information
• Visual Inspection
• Products, parts, components, …
• Process monitoring
• Sound Analytics
• Car engines, bearings, …
• Assembly line inspection
• Data Insights
• Predictive maintenance
• Identify risks
• Optimize operations
• Find dependencies
• Avoid downtimes
➢ Ensure High Quality
Standards

Cognitive Services – key features
1. Visual Inspection
– Inspection of products, parts, components,
and more.
– Continuous process monitoring for quality
assurance.
2. Sound Analytics
– Analyzing sound data for detecting issues in car
engines, bearings, and across assembly lines.
3. Data Insights
– Facilitating predictive maintenance to prevent
unexpected downtimes.
– Identifying and mitigating risks proactively.
– Optimizing manufacturing operations through
data-driven insights.
– Discovering dependencies within
your production processes.
4. High-Quality Standards Assurance
– Ensuring adherence to stringent quality
standards throughout production.
5. Real-time Alerts
– Providing instant notifications for critical issues,
enabling swift action.
6. Enhanced Efficiency
– Streamlining operations and reducing waste
through data-driven optimization.
7. Customization and Scalability
– Tailoring AI models to your specific
manufacturing needs.
– Scaling the solution as your production
demands grow.
INFORMATION
AI

Visual Inspection - manufacturing
DATA
EDGE
COMPUTE
Lenovo
ThinkEdge
SE50
INFORMATION
AI
Alarm
✓ Automate quality inspection
✓ Small form factor
✓ Operating in difficult environments
✓ Wireless options
✓ Scalable architecture
Other
options
available

Visual Inspection – paper industry
DATA
EDGE
COMPUTE
Lenovo
ThinkEdge
SE50
INFORMATION
AI
Alarm
Other
options
available
Beginning of line 6000mm
End of line 8000mm
Headbox
Press
Section
Image Credit: PIV measurements of flow through forming fabrics: researchgate.net/figure/Forming-section-of-a-modern-Fourdrinier-paper-machine_fig3_286055124
Surface Water Wet Line
✓ Continuous monitoring, 24/7/365
➢ Detect, measure, and analyze wet line

Sound analytics – automotive
DATA
EDGE
COMPUTE
Lenovo
ThinkEdge
SE50
INFORMATION
AI
Alarm
Other
options
available
✓ Consistent and reliable results
✓ Cost-effective, offloads humans
✓ Eliminates potential human errors
✓ 24/7, continuous quality monitoring
✓ Increases overall reliability of production

Data Insights
DATA
(EDGE)
COMPUTE
Lenovo
ThinkEdge
SE50
INFORMATION
AI
Summary
SE350
SE450
Other
options
available
✓ Seamlessly turn data from various sources
(IoT sensors, documents, online weather forecasts, etc.)
into actionable information for better decisions
✓ Find answers hidden in the data
(optimize operations, find dependencies, …)
✓ Enable advanced predictive maintenance
✓ Detect risky situations
SENSORS,
HISTORIC
DATA,
ONLINE DATA,
OTHER
SOURCES

Optimized Maintenance
24
Predictive Reactive Preventive
Timing When required After breakdown At predefined intervals
Pros
Lowest risk of
breakdown
No fixed costs
Lower risk of breakdown
than reactive
Cons High fixed cost Higher risk of breakdown Unnecessary maintenance
Image Credit: Reactive Maintenance Vs Preventive Maintenance Vs Predictive Maintenance (assetinfinity.com)

Performance & Scalability
camera feeds
ThinkSystem SR650V2
50-150
ThinkEdge SE350
5-20
ThinkEdge SE450
10-100
ThinkEdge SE50
1-5
ThinkSystem SR670V2
150+
AI
Always optimized for the latest
hardware available.
Up to 25x faster
Edge AI
(*) Performance enhancement achieved through the optimization of byteLAKE's Cognitive Services
with Intel® OpenVINO™, Intel® DL Boost Vector Neural Network Instructions (VNNI), and others.
Cost
Efficient
Scalable
Solution
Maximum
Performance
Quick deployment
& no external dependencies.
On-premises
Example configurations.
Other options available.
Hardware

Visual Inspection
Automated visual
inspection & objects
recognition
• scratches
• cracks
• dents
• wrong color
• paint chips/peeling
• wrong shape
• fractures
• count objects
• read and analyze
labels
• monitor production
processes and
visually detect
anomalies
• etc.
Sound Analytics
Automated quality
inspection based
on sound analysis
• enable proactive
maintenance
• car engines quality
check
• monitor bearings
performance
• inspect assembly
lines
• analyze sound
samples, filter out
noise, identify
characteristic parts
• detect anomalies
• etc.
AI solving problems
Data Insights
Converting DATA
into actionable
INSIGHTS
• understand why
something happens
• what will likely
happen and when?
• find optimal
configuration
• explore
dependencies
• etc.

Unlocking Opportunities with AI
Anomaly Detection, Reduced Downtime, Increased Productivity
~80%
Companies affected by
unexpected downtime
in the past 3-4 years.
$200K
Average cost of unexpected
downtime per hour
~4 Hrs
Average equipment
breakdown time.
Average loss: $1,040,000.
25%
Lower
maintenance costs
70%
Less
breakdowns
35%
Less
downtime
>75%
Zero unexpected downtime
as a top priority
for most organizations.
+20%
Increased
productivity
The average cost of an unplanned downtime is USD $220,000 a day for a paper or pulp plant.
International Journal of Strategic Engineering Asset Management
Too Early / Too Late
Maintenance
Unexpected
Downtime
improving
over time

• Accelerate Data Analytics
– Processing data from various sources, including images, videos, and sensors.
• Automate Quality Inspection
– Ensuring high accuracy in inspecting products and processes.
– Eliminating potential human errors for consistent and reliable results.
– Increasing overall quality and reliability.
• Optimize Operations and Maintenance
– Reducing unnecessary inspections and lowering maintenance costs.
– Predicting potential failures and downtimes.
• Continuous Monitoring
– Offering 24/7/365 monitoring without boredom or distraction.
– Offloading and supporting human operators.
• Easy Replication
– Enabling quick deployment.
– Functioning offline without an internet connection.
• Continuous Improvement
– The solution can learn and improve over time.
Benefits
offered by byteLAKE’s Cognitive Services
~80%
Human quality control's
average accuracy
99%
Accuracy with AI
Growing AI
adoption
2019
2020
>80%
<50%

AI in Manufacturing - benefits
• Enhanced Productivity
– Streamlining processes for increased productivity.
– Efficient resource allocation based on real-time data.
• Customization and Adaptability
– Tailoring AI models to specific manufacturing requirements.
– Adapting to changing production needs seamlessly.
• Reduced Downtime
– Minimizing production downtime through predictive maintenance.
– Optimizing machine uptime and reliability.
• Data-Driven Decision-Making
– Empowering decision-makers with actionable insights.
– Enabling data-driven strategies for process improvement.
• Consistent Quality Control Across the Organization
– Ensuring consistent product quality throughout the production process.
– Meeting industry standards and regulations effortlessly.
AI

Evolution of Industry: 4.0 and beyond
Industry 1.0
Mechanization
Steam power
Weaving loom
Industry 2.0
Electricity
Mass production
Assembly line
Industry 3.0
Computers and
electronics
Automation
Industry 4.0
Cyber Physical
Systems
Internet of tings,
Networking,
Big Data,
Artificial
Intelligence
Industry 5.0
Human-centric
and resilient
European
industry.
Reinforces the role
and the
contribution of
industry to society
1760-1840
1830s-1915
1960-2010
2011 - Today
2020 - ????

Case Study – AI for Industry 4.0
✓ High Performance and Accuracy
✓ Edge AI Optimized Solution
✓ Continuous Monitoring
✓ Industry 4.0 Automation
Automated Quality Inspection

Cognitive Services
for Restaurants
Simplify Restaurant and Retail operations
with our efficient self-checkout solution,
recognizing meals and groceries
and elevating customer satisfaction.

Self-checkout & Object Recognition
• Self-Checkout:
• Camera recognizes meals
and groceries
• Recognized items are sent
to the POS terminal
• Complete Solution:
• Integration with Simpra
• Includes POS, payment
functionalities, and more
• Easy Installation:
• Requires PC and camera
• Simpra + POS / integration
➢ Enhances Customer
Satisfaction

PICTURE
Self-checkout for Restaurants
PC + POS
LIST
OF
ITEMS
AI
Check created
automatically
1. Place items on a tray
2. Approach the cashier
3. The cashier taps on "Check Create with AI"
4. A picture is taken, and AI recognizes the items
5. The list of items is sent to the POS
6. List can be reviewed / updated
7. Payment
PAYMENT

Integrated with Simpra
New Generation Restaurant, Hotel and Payment Solutions
• Key Features
• POS System
• Stock Tracking Program
• Customer Loyalty Program
• QR Menu Order
• Mobile Reporting
• Table Reservation System
• Kitchen Display System
• Waiter Order Module
• Delivery Service
• Various Integrations

Benefits
offered by byteLAKE’s Cognitive Services for Restaurants
1. Payment Automation
– AI-powered self-checkout.
– Option for manual review and updates.
2. Enhanced Customer Satisfaction
– Shortened queues and wait times.
– High accuracy with close to zero errors.
3. Innovative Solutions
– Brings innovation to every restaurant.
– Integrated with Simpra.
– Available as a complete solution, including
payment.
4. Flexibility and Integration
– Compatible with various POS terminals.
– Option to integrate with local services.
– Easy expansion of AI capabilities for different
retail scenarios.
5. Stability and Performance
– High performance and stability guaranteed.
– 100% offline functionality with no Internet
connection required.
6. Cost-Efficiency
– Transparent, flat-rate pricing.
– No hidden costs or external dependencies.
7. Quick Deployment
– Faster service leads to increased sales
and revenue.
– Reduces manual work and operational costs.

Case Study
✓ Enabled efficient self-checkout
✓ Reduced queues and wait times
✓ Improved customer satisfaction

byteLAKE’s
Cognitive Services
for Restaurants
Optimized for technologies
byteLAKE.com/en/AI4Restaurants Product co-financed by

Deployment
& Licensing
byteLAKE’s Cognitive Services

Cognitive Services
Deployment Architecture
AI INFERENCE
DATA CAPTURE &
PREPROCESSING
AI TRAINING
AI
Updated AI Model
Predicted Results
Real-Time Data for Analysis
Data
Time AI
AI Model
Information
Updated AI Model
Data
Collection
Extracting,
storing, retrieving
Data
Cleaning
Cleaning,
formatting,
labeling
AI Model
Training
Calibration
AI Solution
Deployment
and
Integration
Actions & Decisions

Cognitive Services
AI Quality Inspection System / Data Insights
• Quality Analytics
• Data Analytics
• Dashboard
• Actions & Decisions
• Production Line
• Assembly Line
• Infrastructure
EDGE AI FRONT-END
Data Information
AI Model
AI

Self-checkout
• Items Recognized by AI
• Check Created Automatically
• Payment
• Items on Tray
• Camera Takes Picture
EDGE AI POS
Data Information
AI Model
AI
PAYMENT
INVENTORY

• Real-time Decision-Making
– Immediate AI analysis at the edge enables rapid decision-making without relying
on external services, critical for industrial applications.
• Energy Efficiency
– Edge AI can optimize energy consumption by processing data locally
and reducing the need for constant data transmission.
• Offline Operation
– Edge AI allows devices to continue functioning and making decisions
even when there is no internet connectivity.
• Redundancy and Reliability
– Distributed edge AI systems can offer redundancy and fault tolerance, ensuring
continued operation in case of device or network failures.
• Enhanced Privacy and Security
– AI processing on the edge device reduces the need to transmit sensitive data
to external servers, enhancing data privacy and security.
• Low Bandwidth Requirements
– Edge AI minimizes the need for continuous high-bandwidth data transfer,
reducing network congestion and associated costs.
• Customization and Adaptation
– Edge AI models can be tailored to specific device requirements
and updated easily to adapt to changing conditions.
Edge AI
Data processed close to where it is produced, on-premises

1. Define the Scenario
– Determine the purpose of AI analytics, such as detecting surface anomalies,
monitoring assembly lines, automating operations, accelerating processes (e.g.,
production, purchasing), automating repetitive and dangerous tasks, improving
analytics for better decision-making, and enabling predictive maintenance. Are
there any other specific objectives?
2. Explain Expectations
– Discuss the placement and quantity of cameras or sensors to be used if known.
– Describe existing and potential data sources, including future ones if known.
– Specify the desired level of accuracy.
– Provide information about production rates and expected system performance,
considering future growth.
– Define uptime requirements and any other specific system requirements.
How to Start - preparations
Cognitive Services

1. Scenario Explanation
– Provide example pictures, videos, or other relevant data.
– Conduct online consultations or arrange in-person meetings as needed.
2. Initial Data Insights
– Explain your data, including types, ranges, and dependencies.
– Identify unusual scenarios or exceptions.
– Determine if historic data is available and note any gaps.
– Discuss data storage methods and assess the need for changes or improvements.
– Share sample data with us.
3. Online Q&A Session
– Conduct an online Q&A session to address questions about the presented data and scenario.
4. Deployment Plan and Schedule
– Present a detailed deployment plan and schedule prepared by byteLAKE.
How to Start - first steps in the project
Cognitive Services

Licensing & Cost of Deployment
Cognitive Services
• Licensing
– Annual/monthly licensing plans for Cognitive Services, including upgrades, customer
care, and support.
• AI Model Development
– Costs for AI model training and calibration.
• Data Management
– Expenses related to data collection and cleaning.
• Hardware and Software
– Hardware costs, including PCs and sensors, as well as any associated licenses.
– Installation expenses.
• Integration and Deployment
– Integration efforts as required for successful deployment.

Learn more
byteLAKE’s Cognitive Services
Blog post series
Website:
byteLAKE.com/en/CognitiveServices-toc
byteLAKE.com/en/CognitiveServices
Contact us
CognitiveServices@byteLAKE.com
INFORMATION
AI

CFD Suite
AI-accelerated Computational Fluid Dynamics
Accelerate your CFD simulations by leveraging the speed and efficiency of artificial intelligence.
Slash simulation times, minimize trial-and-error costs, and supercharge decision-making
for heightened productivity.

CFD, Computational Fluid Dynamics
Solving exciting problems across industries
Fire spread simulation
Pollution spread simulation
Aerodynamic simulation
Ensure efficient flow Improve taste

Perform CFD Analysis
to check the solids suspension profile
CFD Simulation
Chemical Mixing
It takes 4-8hrs to complete such simulations. AI can reduce that time to minutes.
Simulations powered by:

AI for CFD?
• Hardware Advancements
– In the past, simulations ran on a few nodes, while today, they can utilize hundreds of nodes.
– Modern processors are significantly faster, accelerating simulation tasks.
• Software Options
– Commercial tools are available, and there are open-source alternatives for various applications.
– Traditional solvers, sometimes with hardware-optimized algorithms, are also in use.
• Turnaround Time Challenge
– Simulations still take days to complete, but customer expectations have evolved.
– Simple flow problems can now be solved within hours,
but customers often expect results within minutes.
• Thinking Outside the Box
– Merely adding more compute power isn't always the solution.
– Considering alternative choices, such as different numerical methods, while addressing concerns about
accuracy.
• Exploring AI Solutions
– Delving into Artificial Intelligence, Deep Learning, and Machine Learning as potential solutions.

AI-accelerated CFD Simulations
byteLAKE’s CFD Suite
• Traditional workflow
• byteLAKE’s CFD Suite
CFD Simulation (iterations, time steps) RESULTS
Visualization
Modeling
RESULTS
Visualization
Hours, weeks
Modeling CFD Suite (AI-accelerated CFD)
2x, 10x, 20x, 40x, …
FASTER time to INSIGHTS
CFD Suite
Collection of innovative AI Models
for computational fluid dynamics.
byteLAKE.com/en/CFDSuite

AI-accelerated CFD, example results
Simulation time reduced: from hours to minutes
Configuration: chemical mixing, <2M cells, 3D data, steady-state, 5K iterations
CFD Solver CFD Suite (AI predictions)
X-Plane
Velocity
X-Plane
Pressure
X-Plane
Turbulent
kinetic
energy
CFD Suite
Simulations powered by:
Example results for: CFD/chemical mixing case study.
Note: accuracy is configurable and depends on requirements.

• Traditional workflow
• byteLAKE’s CFD Suite
Simulate RESULTS
Visualization
Modeling
RESULTS
Visualization
Hours to Weeks
Modeling Simulate
2x, 10x, 20x, 40x, …
Accuracy is configurable and depends on requirements.
Simulate
AI supervisor recognizes the data pattern
and replaces the part of the CFD simulation with a prediction
How does it work?
Faster Time to Insights
(2x, 10x, 20x, 40x, …)
with byteLAKE’s
CFD Suite’s Learning-on-the-Fly
AI Models

Modeling
AI Supervisor
RESULTS
Visualization
Simulate Simulate
Simulate Simulate
AI Accelerator AI Accelerator AI Accelerator
CFD Suite’ collaborating modules generate results
• AI Accelerator, guarantees acceleration and makes predictions based on a trained pattern
• AI Supervisor, guarantees accuracy and decides to:
– Accelerate once or multiple times during simulation
– Stop the simulation and return the physics-aware results
CFD Suite
AI Accelerator and AI Supervisor
2x, 10x, 20x, 40x, …

AI model training & calibration
AI
Past simulations
(historic data)
CFD Suite
AI
➢ Leverage data generated
by past simulations
➢ Learning-on-the-Fly

Hardware
Ensuring high portability across various hardware
and operating system configurations.
CPUs GPUs TF ready
Linux OS Ubuntu CentOS RedHat
CFD Solver Providers CFD Solver A CFD Solver N…
AI Trainer
Keras Horovod
TensorFlow (TF)
CFD AI Connector C++
Python Intel OpenVino
Other hardware accelerators, compatible with TensorFlow.
(*)
(*)
Software Stack (AI training)

AI Predictions
INPUT
DATA
ON-PREMISES
AI-
ACCELERATED
CFD
AI
RESULTS
Various
options
available 3D DATA

Ensuring high portability across various hardware
and operating system configurations.
Hardware
OS
CFD Solver Providers CFD Solver A CFD Solver N…
AI Accelerator
AI Supervisor
(AI Trainer)
Keras
TensorFlow (TF)
CFD AI Connector C++
Python Intel OpenVino
Ubuntu CentOS RedHat Windows
CPUs GPUs
Software Stack (AI predictions / deployment)

AI
Always optimized for the latest hardware available.
AI training optimized for edge servers and multi-node HPC architectures.
Edge Servers and HPC
Cost
Efficient
Scalable
Solution
Maximum
Performance
Quick deployment
On-premises
Hardware
ThinkSystem SR650V2
ThinkEdge SE350 ThinkEdge SE450
Workstations ThinkSystem SR670V2
AI Predictions AI Training

Benefits
offered by byteLAKE’s CFD Suite
Faster Time to Insights
– Swift simulation results enable quicker decision-
making and problem-solving.
Cost Reduction
– Lower costs associated with reduced trial
and error experimentation.
Rapid Design Iteration
– Accelerated simulations allow
for faster prototype design and testing
Improved Productivity
– Enhanced efficiency in research
and development processes.
Enhanced Safety Measures
– Quick assessments of safety protocols
and potential risks.
Energy Efficiency Optimization
– Faster insights into optimizing energy
consumption and resource utilization.
Resource Conservation
– Reduced resource consumption
in experimental setups.
– Leverage data generated by past simulations
Competitive Advantage
– Faster product development
and innovation lead to a competitive edge.
Real-time Monitoring
– Possibility of real-time monitoring
for immediate adjustments.

CFD Suite
– case study
AI-accelerated Computational Fluid Dynamics
Accelerate your CFD simulations by leveraging the speed and efficiency of artificial intelligence.
Slash simulation times, minimize trial-and-error costs, and supercharge decision-making
for heightened productivity.

• Many everyday products start in small-scale
settings, like home labs, where unique recipes
are crafted in pots and pans.
• Scaling up production demands larger tanks.
• The quality of mixing during manufacturing
directly influences product quality.
➢CFD Simulations play a crucial role.
Background: chemical mixing

• We've chosen a specific phenomenon for benchmarking,
aiming to calculate the stable state of a liquid mixture in
a tank featuring a single impeller and baffles.
• By adjusting input parameters,
we simulate several quantities:
– Velocity vector field (U)
– Pressure scalar field (p)
– Turbulent kinetic energy (k)
– Turbulent dynamic viscosity (mut)
– Turbulent kinetic energy dissipation rate (epsilon)
Scenario: chemical mixing
3D DATA

AI Predictions
AI-
ACCELERATED
CFD
CFD Suite
AI
Results generated by CFD Suite
Results generated by CFD Solver

Panel Discussion:
CFD Suite accelerating Chemical Mixing
✓ Trained for expediting
chemical mixing
simulations.
✓ Optimized for a range
of hardware options.
✓ Versatile and capable
of training with different
CFD simulation types.

Among 5 top startups working on CFD!
“Explore our analysis of 441
global startups & scaleups
and learn how their
computational fluid dynamics
(CFD) solutions impact your
business!”
“This time, you get to discover
5 hand-picked startups
developing computational
fluid dynamics solutions.”

Deployment
& Licensing

1. Define the Scenario
– Identify the target CFD solver for acceleration and explain associated processes
and scenarios (parameters, ranges, dependencies, geometries, etc.).
2. Explain Expectations
– Define the required accuracy levels.
– Specify supported input configuration ranges.
– Share insights on anticipated system performance, including future scalability.
– Outline integration needs and the desired interaction of CFD Suite with other tools
(data formats, API, etc.).
How to Start - preparations

1. Scenario Explanation
– Provide example data.
– Conduct online consultations or arrange in-person meetings as needed.
2. Initial Data Insights
– Explain your data, including types, ranges, and dependencies.
– Identify unusual scenarios or exceptions.
– Determine if historic data is available and note any gaps.
– Discuss data storage methods and assess the need for changes or improvements.
– Share sample data with us.
3. Online Q&A Session
– Conduct an online Q&A session to address questions about the presented data and scenario.
4. Deployment Plan and Schedule
– Present a detailed deployment plan and schedule prepared by byteLAKE.
How to Start - first steps in the project

Licensing & Cost of Deployment
• Licensing
– Annual/monthly licensing plans for CFD Suite, including upgrades, customer care,
and support.
• AI Model Development
– Costs for AI model training and calibration.
• Data Management
– Expenses related to data collection and cleaning.
• Hardware and Software (if needed)
– Hardware costs, as well as any associated licenses.
– Installation expenses.
• Integration and Deployment
– Integration efforts as required for successful deployment.

Join the CFD Suite community
Blog post series,
discussions forum etc.

Links
79
LinkedIn Group
Facebook Group
Blog post series
Website
bytelake.com/en/CFDSuite-FB-group
bytelake.com/en/CFDSuite-LN-group
bytelake.com/en/AI4CFD-toc
bytelake.com/en/CFDSuite
CFDSuite.com

AI-
accelerated
CFD
CFD Suite

Meet byteLAKE
AI Solutions for Industries |
Quality Inspection |
Data Insights |
AI-accelerated CFD |
Self-Checkout
Products:
CFD Suite Cognitive Services
www.byteLAKE.com
Headquartered in Poland
Empowering Industries with Artificial Intelligence
Solutions.
At byteLAKE, we harness cutting-edge technology to
provide advanced quality inspection and data insights
tailored for the Manufacturing, Automotive, Paper,
Chemical, and Energy sectors.
Additionally, we offer self-checkout stations for
Restaurants and object recognition solutions for Retail
businesses.
+48 508 091 885
+48 505 322 282

byteLAKE
AI Solutions
Manufacturing, Automotive, Paper,
Chemical, and Energy sectors
Partnership with Intel

byteLAKE @ Intel
• RRK / RFP Ready Kit
➢ Storefront: https://www.intel.com/content/www/us/en/partner/showcase/storefront/a5S3b0000002k8EEAQ/bytelake.html
➢ Playbook: Intel® IoT RFP Ready Kit Solutions Playbook (Jan’23 edition, page 62)
• AI in Production
➢ https://www.intel.com/content/www/us/en/internet-of-things/ai-in-production/partners/bytelake.html
• AI Builders
➢ Intel® AI Builders - Membership Partners (search: byteLAKE)

Intel & byteLAKE publications
• Intel AI
https://twitter.com/IntelAI/status/1590751152315854854?s=20&t=iKEyWR9qYB7cqXlm32QkWw
(joint promotion of byteLAKE’s CFD Suite)
• Intel’s newsroom
https://www.intel.com/content/www/us/en/newsroom/news/ai-helps-speed-papermaking-process-europe.html
(Cognitive Services for Paper Industry)
➢ Follow-up article: https://www.insight.tech/retail/paper-mills-press-on-with-ai-visual-inspection
• Intel Software / Inside AI People series
https://youtu.be/bUt9J7FMyVs
• Case Study
https://www.intel.com/content/www/us/en/data-center/idc-bytelake-case-study.html
➢ Follow-up article: https://www.insight.tech/industry/applying-industrial-ai-models-to-product-quality-inspection

Products optimized for Intel technologies

AI
Always optimized for the latest
hardware available.
Up to 25x faster
Edge AI
(*) Performance enhancement achieved through the optimization of byteLAKE's Cognitive Services
with Intel® OpenVINO™, Intel® DL Boost Vector Neural Network Instructions (VNNI), and others.
Maximum
Performance
AI-
accelerated
CFD

Contact us to learn more!
mkolanko@byteLAKE.com
+48 505 322 282
Mariusz Kolanko
mrojek@byteLAKE.com
+8 508 091 885
Marcin Rojek

NVIDIA Expertise
AI Solutions for Industries

• In-depth understanding of NVIDIA hardware and software ecosystems:
– Expertise in various NVIDIA architectures, including Fermi, Kepler (K80, GeForce GTX Titan,
Jetson), Maxwell (NVIDIA GeForce GTX 980), Pascal (P100), Tesla V100/A100/H100, and T4.
– Mastery of GPU programming languages such as CUDA, OpenCL, and OpenACC.
– Extensive experience optimizing solutions for NVIDIA GPUs.
– Comprehensive knowledge spanning desktop, mobile, and server environments.
• Demonstrated success through multiple case studies:
– AI training (machine learning and deep learning)
– Edge AI inferencing
– Classic HPC simulations (Computational Fluid Dynamics, weather simulations)
• Active participation in cutting-edge research with numerous publications in prestigious
journals, including Concurrency and Computation: Practice and Experience, Parallel
Computing, and the Journal of Supercomputing.
Expertise Across NVIDIA Architectures
and Configurations

AI
Always optimized for the latest hardware available.
AI training optimized for edge servers and multi-node HPC architectures.
Edge Servers and HPC
Cost
Efficient
Scalable
Solution
Maximum
Performance
Quick deployment
On-premises
Hardware
ThinkSystem SR650V2
ThinkEdge SE350 ThinkEdge SE450
Workstations ThinkSystem SR670V2
AI Predictions AI Training
Edge HPC Server Performance: AI training
➢ Lenovo’s ThinkEdge SE450
➢ 2 NVIDIA A100 80GB Tensor Core GPUs
➢ Intel Xeon Gold 6330N CPU
➢ Download the full report

HPC Simulations optimized by Machine Learning
automatic adaptation of algorithm to a specific hardware architecture
• Enables software portability between different architectures:
– CPU: different number of cores, hierarchy of memory, caches size;
– GPU: register file reusing, shared memory utilization, GPU direct support, reduction of global memory transaction;
– HPC: selecting the right number of nodes, scalability estimation, overlapping data transfers & communication;
– Hybrid: load balancing
(i.e. selecting appropriate parts of the algorithm for different devices executing the code with different performance)
• Helps build adaptable algorithms:
– automatically selecting the size of data blocks, number of threads, number of processes, precision of data
(i.e. depending on algorithm or input data characteristics some data can be stored using double, single of half precision format)
– selecting the criterion of optimization: performance, energy consumption, accuracy of result, mix
(i.e. mix of performance & energy to minimize energy and keep execution time or to optimization performance & keep energy budget)
• Can auto-configure the system and provide the most suitable compiler flags
byteLAKE’s Software Autotuning

• Goal: Reducing the energy consumption of the MPDATA algorithm (algorithm for numerical simulation of
geophysical fluids flows on micro-to-planetary scales – especially used in a numerical weather prediction).
• Hardware: Piz Daint supercomputer (ranked 3-rd at top 500), equipped with the most advanced Pascal-based
GPUs: NVIDIA Tesla P100.
• Idea: Applying mixed precision arithmetic - set a part of operations to be performed in a single precision (32-bits)
and the remaining set to double (64-bits).
• Why do we use it? A single simulation of the weather phenomenon needed more than 1013
operations. We
suspected that not all of them needs double precision arithmetic to preserve the same simulation accuracy. We
believe that the control of the precision and accuracy of numerical results can increase the performance, decrease
the energy consumption, and provide highly accurate results.
• Solution: We used unsupervised learning to estimate the correlation between the precision of each matrix and
their influence on criteria (energy, accuracy of results). During the dynamic and short training stage we evaluated
the set of operations that could be performed in a single precision without loss in accuracy of the weather
simulation.
Research Case: (concluded)
CFD acceleration with GPU (MPDATA, weather forecast)
Results: We reduced energy by 33%, increased
performance by the factor of 1.27x using 25% less GPUs,
keeping the accuracy of the results at the same level
as when using double precision arithmetic.

Research Case:
Reconfiguring HPC Simulation with AI to optimize performance and energy
node count
accelerators per node
memory alignment
streams count
buffering types
…
cpu cores
memory policy
1 000 000 000
Possible
configurations
Ca. 5000
possible
configurations
Artificial Intelligenc
e
This module utilizes among
others the supervised learning
method with the random forest
algorithm.
The main functionality of the module
is to prune the search space in
order to eliminate the worst
configurations.
We develop a Machine
Learning module in order to
select the most fitting
configuration.
In this way we achieve a small set
that at 90% contains the best
configuration.

MPDATA Accelerated
CFD / Advection algorithm optimized for heterogeneous
computing. For Nvidia GPUs and Xilinx Alveo FPGAs.

• MPDATA
(Multidimensional Positive Definite Advection Transport Algorithm)
– main part of the dynamic core of the Eulerian/
semi-Lagrangian (EULAG) model
– EULAG (MPDATA+elliptic solver) is the established computational model,
developed for simulating thermo-fluid flows across a wide range of scales
and physical scenarios
– currently, this model is being implemented as the new dynamic core of the COSMO
(Consortium for Small-scale Modeling) weather prediction framework
– advection (together with the elliptic solver) is a key part of many frameworks that allow
users to implement their simulations
• Advection
– movement of some material (dissolved or suspended) in the fluid.
Algorithm: Advection (MPDATA)
General Information

• Easy to integrate
– Can work as a standalone application or be called as a function via our dedicated interface
(e.g. can be called as a function with input and output arrays)
– Compatible with frameworks like TensorFlow for integrating deep learning with CFD codes
• Easy to visualize the results
– Results can be stored in a raw format as a binary file of the output arrays or converted via
byteLAKE tools to a ParaView format
• See benefits already in 1-node HPC configurations
– Strongly adapted to Alveo U250, were single card supports the max size of arrays: 2,1 Gcells
(max compute domain: 1264 x 1264 x 1264) ~ 60 GB
• Scalable to many cards per node and many nodes
byteLAKE’s implementation compatibility

• First-order-accurate step of the advection scheme.
Second-order is an option.
• Input data
– Array X – non-diffusive quantity
(e.g. temperature of water vapor, ice, precipitation, etc.)
– Arrays V1, V2, V3 - each of them stores the velocity vectors in one direction
– (optional) Arrays Fi, Fe - implosion and explosion forces acting on a structure of X
– (optional) Array D with density
– (optional) Array rho which defines an interface for the coupling of COSMO and EULAG dynamic core
(used to provide the transformation of the X variable)
– DT – time step (scalar)
• Output data
– single X array that was updated in the given time step
Technical Information

• Applications include
– To characterize the sub-grid scales effect in global numerical simulations
of turbulent stellar interiors
– To compare anelastic and compressible convection-permitting weather forecasts
for the Alpine region
– Modeling the prediction of forest fire spread
– Flood simulations
– Biomechanical modeling of brain injuries within the Voigt model
(a linear system of differential equations where the motion of the brain tissue depends
merely on the balance between viscous and elastic forces)
– Simulation gravity wave turbulence in the Earth's atmosphere
– Simulation of geophysical turbulence in the Earth's atmosphere
– Ocean modeling: simulation of three-dimensional solitary wave generation and
propagation using EULAG coupled to the barotropic NCOM (Navy Coastal Ocean Model)
tidal model
101
Applications of Advection (MPDATA)

• Applications include cont.
– Oil and Gas: provides a significant return on investment (ROI) in seismic analysis,
reservoir modelling and basin modelling. Used also to monitor drilling and seismic data
to optimize drilling trajectories and minimize environmental risk.
– AgriTech: models to track and predict various environmental impacts on crop yield such
as weather changes. For example, daily weather predictions can be customized based on
the needs of each client and range from hyperlocal to global.
• Example adopters
– Poznan Supercomputing and Networking Center, Poland: prognosis of air pollution
– European Centre for Medium-Range Weather Forecasts, UK: weather forecast
– Institute of Meteorology and Water Management, Poland: weather forecasts
– German Aerospace Center: aeronautics, transport and energy areas
– University of Cape Town, RPA: weather simulation
– Montreal University: weather simulation
– Warsaw University: ocean simulation
Applications of Advection (MPDATA), cont.
Full list

12x better performance
30% reduced energy consumption
• Our solution: machine learning managed, dynamic application of mixed precision
• Highlights:
– Dynamic estimation of the algorithm’s power consumption as a function of the
frequency of the processor and the number of cores.
– Energy-aware task management
– Auto-tuning procedure taking into account algorithm’s and GPU-specific parameters
for auto-configuring purposes.
– Result: better performance, less energy consumed.
Weather engine optimized for Europe’s fastest
supercomputer (Piz Daint)
Our mechanism provides the energy savings of up to 1.43x
comparing to the default Linux scaling governor.

Dynamic Mixed Precision, cont.
We reduced E by 33%, increased performance by the factor of 1.27x using 25% less GPUs.
We kept the accuracy of the results at the same level as when using double precision arithmetic.

Dynamic Mixed Precision
Optimize execution time
• Ported geophysical model
(EULAG) to a parallel
computing supercomputer
architecture (Piz Daint)
• Used Machine Learning
(Random Forest) to
optimize various numerical
parameters as: data blocks
sizes, number of GPU
streams, sizes of vector data
types
Optimize energy efficiency
• Developed a mechanism
(mixed precision) that
allowed for providing a low
energy consumption of
supercomputers keeping the
code performance at the
highest possible level
• Developed a framework,
based on software
automatic tuning approach
Results
✓ 10 times faster
✓ Then we improved it
even more, reaching the
speed-up of 1.27
✓ Energy consumption
reduced by 33%
✓ Optimized GPUs usage
while keeping the accuracy
of computations
Highlights:
• C++, CUDA, MPI, OpenMP

Expertise Software Services
Alveo Solutions

Expertise in Alveo FPGA programming
PCIe
x86 CPU
Host
Application
Runtime and Drivers
Acceleration API
FPGA
Accelerated
Functions
DMA Engine
AXI Interfaces
byteLAKE’s
Solutions
Xilinx
Acceleration
Platform
C/C++ code
with
OpenCL API calls
C/C++
or
OpenCL C
FPGA
CPU
107

• Xilinx pioneered C to FPGA compilation technology (aka “HLS”) in 2011
108
Source code in C, C++ or OpenCL
loop_main:for(int j=0;j<NUM_SIMGROUPS;j+=2) {
loop_share:for(uint k=0;k<NUM_SIMS;k++) {
loop_parallel:for(int i=0;i<NUM_RNGS;i++) {
mt_rng[i].BOX_MULLER(&num1[i][k],&num2[i][k],ratio4,ratio3);
float payoff1 = expf(num1[i][k])-1.0f;
float payoff2 = expf(num2[i][k])-1.0f;
if(num1[i][k]>0.0f)
pCall1[i][k]+= payoff1;
else
pPut1[i][k]-=payoff1;
if(num2[i][k]>0.0f)
pCall2[i][k]+=payoff2;
else
pPut2[i][k]-=payoff2;
}
}
}
FPGA
Compile

• Compute domain divided
into sub-domains
• Host sends data to the FPGA
global memory
• Host calls kernels to execute them on
FPGA (kernel is called many times)
• Each kernel call represents
a single time step
• FPGA sends the output array
back to the host
Typical Architecture

• Kernel is distributed
into 4 SLRs
• Each sub-domain is
allocated in different
memory bank
• Data transfer occurs
between neighboring
memory banks
Example processing
SLR0
Kernel_A
SLR1
Kernel_B
SLR2
Kernel_C
SLR3
Kernel_D
Kernel
Bank0 Bank1
Bank2 Bank3
Sub-domain Sub-domain
Sub-domain Sub-domain
19

Case study: CFD Kernels adaptation
Typical CFD workflow
From CAD to MESH…
(meshing)
Image source: https://www.openfoam.com/products/visualcfd.php
…to CFD simulation and visualization.
• MESH conversion (input)
• byteLAKE’s CFD Kernels
• Data output for visualization
up
to
5%
of
simulation
time
major
workload
OPENFOAM® is a registered trademark of ESI Group. This offering is not approved or endorsed by ESI Group, the producer of the OpenFOAM software and owner of the OPENFOAM® and OpenCFD® trademarks.

byteLAKE created set of highly optimized CFD kernels
for Xilinx Alveo Datacentre accelerator cards
–Advection (movement of some material, dissolved or
suspended in the fluid)
–Pseudo velocity (approximation of the relative velocity)
–Divergence (measures how much of fluid is flowing into/ out
of a certain point in a vector field)
–Thomas algorithm (simplified form of Gaussian
elimination for tridiagonal system of equations)
112
CFD Kernels

CFD acceleration with Alveo FPGA
113

Alveo FPGA Benchmark
INTEL XEON
E5-2995
INTEL XEON
E5-2995
INTEL XEON
GOLD 6148
INTEL XEON
PLATINUM
8168
XILINX
ALVEO U250
Performance (the higher
the better)
INTEL XEON
E5-2995
INTEL XEON
E5-2995
INTEL XEON
GOLD 6148
INTEL XEON
PLATINUM
8168
XILINX
ALVEO U250
Energy (the lower the
better) INTEL XEON
E5-2995
INTEL XEON
E5-2995
INTEL XEON
GOLD 6148
INTEL XEON
PLATINUM
8168
XILINX
ALVEO U250
Performance/W (the
higher the better)

Accelerates and automates document processing.
OCR for invoices, AI powered. No templates!

OCR for invoices, AI-powered. No templates!
Accelerates and automates document processing.
Tangible benefits and ROI thru:
• significantly reduced time of documents processing
• eliminated human error
• Artificial Intelligence (AI) driven automation
Meet
Bottom line:
brainello saves money and takes away boring stuff off your schedule.

➢OCR for invoices, supercharged with AI
(no need to prepare templates)
➢Various formats of documents supported
➢Easy integration
(standalone engine or integrated
into larger system thru many supported interfaces)
➢Self-improving over time
(assisted learning as an option)
Key features

Example Case Study:
Integrated with Bpower2
7xFaster Invoices
Processing
documents
workflow

• Documents processing automation:
121
Beyond invoices automation
Roadmap
– Content Analytics
(finding key parts like dates, regulations and setting
reminders, enforcing rules etc.)
– Document Type Detection
(sorting based on content and distributing
to appropriate workflows)
– Intelligent forms filling
(funds applications and repetitive forms)
FRAMEWORK

In Collaboration with Microsoft

brainello
• Documents processing automation
• 1800 invoices processed / hour !
• No templates needed. AI powered.
• Learning over time

Ewa Guard
AI for Agriculture. Large areas visual analytics.
Earth Water Air

Powered by:
AI for Drones
Large Areas Visual Analytics

Forestry Management with AI
AI analyzing drones’ footage
& reducing manual work in reforestation

Video: byteLAKE.com/en/EwaGuard-video

Detecting young trees
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree
Tree

Ewa Guard’s AI Model Training optimized & accelerated
with Lenovo AI Innovation Centers
• Ewa Guard’s Deep Neural Network
Reduced 336 hours (2 weeks) to 6 hours
56 x speed-up
Layers: 27
Epochs: 10 000
Images: 400 (416x416 px)
Accuracy: 92-95%
Original HW:
2x Intel Xeon E5-2695 CPU
256 GB of RAM
Lenovo HW:
SR650
2x Intel Xeon Gold 6148 CPU
Nvidia T4
768GB of RAM

Illegal Dumping Detection
Video: https://youtu.be/x1IMYfSE9qI

Illegal Dumping Detection in numbers
Solution brief:
• Layers: 32
• Epochs: 15 000
• Images: 650 (416x416 px)
• Accuracy: 90% tyres detected
Hardware:
• NVIDIA GTX 1060
• Learning time: 12h
Layers:
• 23 convolution layers
• 1 region layers
• 2 route layers
• 1 reorg layer

Ewa Guard – official trailer
Cleaning up the world with Artificial Intelligence

Mitigating the Impact of Deforestation
Did you know that Earth loses ~19 million acres of forests per year,
which is equal to 27 soccer fields every minute?

Demo at SC19 in Denver, Colorado
AI & HPC Convergence in Action

Ewa Guard:
1. Integrate Ewa Guard
into your system (on-
premise, Cloud)
2. Ewa Guard will analyze
drones’ footage and
provide analytics
Ewa Guard can be trained
to analyze images or videos
and extract the information
you need.
Contact us to learn more!

Contact us at EwaGuard@byteLAKE.com
to learn more and request a demo.
Learn more:
byteLAKE.com/en/EwaGuard
+48 508 091 885
+48 505 322 282

Federated Learning
Enabling collaborative learning with privacy for Edge AI
applications. Leveraging AI models across distributed
networks.

How it works
Machine
Learning
Sensor
Data
Time
Local training & inferencing
Edge AI
Aggregating local models
(black box)
Federated Learning
Data Center
White Paper

➢ learning to predict pressure changes (locally)
➢ learning locally & from already trained models (federated learning)
Manufacturing
Simulation
Fan 1 Fan 2
Barometric
pressure
sensor
Filter 1 Filter 2
Styrofoam
pellets
Styrofoam pellets…
… cause filters clogging
& air pressure changes.

✓Enables Scalability
(Decentralized AI enables IoT/devices to learn from each other)
✓Solves low-throughput and high-latency challenges
(Local AI models provide real time response and lower power consumption)
✓Improves accuracy
(Have smarter models via aggregation of many local models)
✓Reduces training time
(Benefit from local training and already trained models in the neighborhood)
✓Lowers the cost of training
(Bringing data from all devices is expensive)
✓Ensures privacy
(Sensitive data stays local)
Benefits of
Federated Learning
White Paper

Models
(DNN)
Interaction
Data
Edge AI Solution
Instant Response
Advanced
Analytics
AI for Lenovo’s Smart Edge Devices

✓ byteLAKE’s Cloud Framework
▪ Communication between Edge device and Microsoft Azure
▪ Visualization of gathered information
(reporting, statistics, diagnostics etc.)
Real-Time
Objects Detection
On Edge
Dashboard
Edge AI
Microsoft
Azure
On-device AI-powered Video Analytics

✓ byteLAKE’s low-level C++ mod
(integrates Basler’s cam with NCS SDK)
✓ byteLAKE’s Computer Vision asynchronous model
(enables real-time on-device objects classification)
✓ OpenCV code to visualize the results
Movidius
Neural Compute
Stick
Raspberry Pi
Real time objects
classification
DNN
TensorFlow
Pylon API
Raspbian Stretch OS
NCS SDK
OpenCV Pineapple!
97.9%
Model
On-device Objects Recognition

• Challenge: perform analytics on a Time Series Data
• byteLAKE approach: designed a concept framework that can analyze
time series data from various sensors and become a foundation for
predictive maintenance systems.
• It relies on two
key mechanisms:
– Feedback Loop Control (FLC).
It selects the machine learning
model that fits best.
– Environment Recognition (ER).
It finds the most critical
parameters for a given scenario.
149
Research Case:
Predictive Maintenance for Industry 4.0

Predictive Maintenance: case study
150

HPC at byteLAKE
Accelerating time to results and adapting complex algorithms
to GPU, FPGA, many-CPU architectures.
Unleashing the power:
• selecting the right programming model to a given problem
(task parallelism, data parallelism, mixture of these two)
• providing the right balance between CPUs and GPUs/FPGAs
• optimizing data transfers between host memory and accelerators
• code adaptation to a variety of computing platforms
Bottom line: lowering TCO thru various optimizations
(performance, energy efficiency, accuracy of calculations)
Making the most of the hardware:
• Speedup: accelerating time to results for complex algorithms
• Green Computing: optimizing algorithms to reduce energy consumption
• Scalability: from single nodes to clusters

Dynamic redistribution of HPC resources during the simulation
shrinking or expanding the number of nodes at runtime
What’s the goal?
• maximization of performance of the entire HPC cluster, instead of a single job.
How it works?
• all or a specific set of jobs is considered as the subject of optimization, not only a single job.
• consequently, a single job within a given set can be executed with lower performance, however,
the entire set of jobs will be executed faster or more energy efficiently (depending on selected criterion)
Features:
• can be integrated with SLURM system (accessible from most computing clusters)
• can manage the part of the cluster resources (nodes 0-16)
or the specific tasks submitted to the cluster (algorithm A, B, C)
• needs to be integrated with the algorithm
byteLAKE’s AI powered HPC Scheduler

HPC Scheduler: results
153
byteLAKE’s AI powered HPC Scheduler
• run app0
(16 nodes)
• shrink app0 to 8 nodes and run app1
(16 nodes)
• shrink app1 to 8 nodes and run app2
(8 nodes)
Result:
• 24 nodes utilized
• 3 applications running using 8 nodes, all
running in parallel
• Performance: efficiency of 8 nodes greater
than 16 as 16 is max. scalability
• Each node more utilized
Traditional SLURM
• run app0 (16 nodes)
• wait for (app0)
• wait for (app1)
• wait for (app2)
Result:
• 16 nodes utilized
• 3 series of execution
Setup: 24 nodes available, 3 jobs. Application scalable up to 16 nodes
(using more than 16 nodes results in performance decrease)

Every success starts with a dream.
We help bring those dreams to life.
156
AI Solution
(1)
Understand
(2) Prepare
(3)
Execute
(4) Evaluate

158
Sprint-based Agile Delivery
Backlog
Planning
Execution
Potentially
deliverable product
Inspect
& adapt
We deliver projects
in Agile fashion
• Backlog Building
it’s a “what” part which tells us what we are to achieve
• Planning
it’s a “how” part: tells us how we can achieve the “what”
• Execution
Building a product in time-boxed iterations (sprints)
• Gradual product delivery
Each sprint ends with a delivery of a so-called:
potentially deliverable product
• Inspect & adapt
Reviews, lessons learnt, continuous improvement
➢ Cycle repeats until product is finalized
➢ Sprints can be grouped in milestones
(I, II, III, …)
I V
II III IV
Team
…

Marcin Rojek, co-founder
Marcin Rojek
Co-founder
Co-Founder of byteLAKE, a company specializing in crafting artificial
intelligence solutions for industries including manufacturing, automotive,
paper, chemical, energy, and the restaurant sector.
An enthusiast for merging research and the latest scientific achievements
with concrete business needs. The conversion of data into valuable insights
is an area where Marcin and his team constantly seek inspiration for further
developing the company's products.
What does artificial intelligence mean to Marcin? It's a tool that efficiently
harnesses both historical and real-time data from various sources, such as
cameras, microphones, and sensors. This enables quality inspections of
products, optimization of production processes, and early detection of
potential faults.
Marcin holds extensive international experience gained through
implementing cutting-edge IT technologies in markets across Europe, the
USA, and Asia. Beyond byteLAKE, Marcin's priorities lie in family and sports.

Mariusz Kolanko, co-founder
Mariusz Kolanko
Co-founder
Co-Founder of byteLAKE, with extensive international business experience. A
practitioner of delivering solutions on the global stage.
Responsible for creating and implementing industrial solutions for major
global firms spanning manufacturing, automotive, paper, chemical, energy,
and restaurant industries.
His background in multinational corporations equips him to confront
challenges in running his own business. An enthusiast for artificial
intelligence and a proponent of harnessing its potential in the business
landscape.
A speaker at various engagements and training sessions, both domestically
and internationally, focusing on artificial intelligence. Outside of his
professional life, Mariusz loves traveling and enjoys an active lifestyle in his
leisure time.

byteLAKE’s CTO
Krzysztof Rojek
byteLAKE's CTO. Pivotal link between byteLAKE’s business and the
dynamic domains of research and academia.
Impassioned advocate and catalyst for ideas that incubate in the research
sphere and later find their place in tangible, real-world business
applications.
Krzysztof holds both a PhD and a DSc in Computer Science, with a
specialization in Parallel Computing, GPGPU, self-adaptable codes, and AI
applications.
His contributions have garnered international recognition, marked by
prestigious accolades within the domains of High-Performance Computing
(HPC) and Artificial Intelligence (AI).

• Vast experience in business projects
2008-2010 Research & Engineering in collaboration with IBM
– Background: IBM’s CELL processors became first accelerator for supercomputing architectures.
New mathematical models were needed for software to make the most of new hardware.
– Task: optimize mathematical algorithms for BLAS (Basic Linear Algebra Subprograms) software
package to guarantee maximum utilization of processors’ available performance.
– Result: CELLs-based computing capacity utilized in 99% + optimized memory access management.
2012-2015 Numerical algorithm re-design for a weather forecasting institution
– Background: A weather forecasting institute (IMGW), together with academic institutions, won a
grant for the purpose of renovating their software assets and porting them to the modern hardware.
– Task: Re-design geophysical simulation algorithm (MPDATA) for parallel computing architecture
and adapt it to the CPU+GPU – based supercomputing environment (PizDaint in that case).
– Result: MPDATA algorithms has been completely redesigned, parametrized and ported to the
massively parallel, many-CPU/GPU architectures. Overall the software’s performance has been
increased by 10 times.
byteLAKE’s CTO
Krzysztof Rojek, cont.

• Vast experience in business projects
2017+ Optimization of Geophysical Algorithms
– Background: byteLAKE & Megware want to build an automate algorithm optimization platform
– Task: incorporate AI into boosting performance and decreasing energy consumption to the weather
forecasting model.
– Result: so far designed a mathematical scheduling model, custom-made random forest AI and
dynamic mixed precision arithmetic that boosts performance further by factor of 1.32 (12 times from
initial state) and slashes the energy consumption by 34%.
2017 AI-based model to process time-series data
– Background: A factory in the US wanted to upgrade their predictive maintenance model utilizing
time-series data from various IoT sensors measuring filters’ delta pressure, humidity, dust level etc.
– Task: design a concept model that could perform an analytics on time-series data and point to the
most influencing parameters for a given phenomenon (i.e. filters clogging).
– Result: designed a feedback-loop-control model that incorporated several AI algorithms in order to
extrapolate given parameters based on their historic data and mechanical characteristics. First
concepts with the statistical simulations are currently in the trials.
byteLAKE’s CTO

• Research & Engineering key contributions
– Algorithms parallelization and adaptation to hybrid CPU-GPU platforms
– Machine and Deep Learning in various aspects i.e. created new techniques
for self-adapting source code, multi-objective code optimization
(energy, performance, accuracy), multimedia files analytics, big data analysis etc.
– Adaptive scheduling with online modeling to minimize the energy consumption while
keeping the given time constraints.
• Many international awards, e.g.:
– “Outstanding monograph” about parallel computing, Polish Academy of Sciences
– “Excellent Paper” about algorithms for heterogonous architectures, Hong Kong
– Grant from the IBM Company for Realization of Workshop, dedicated to programming and
usage of advanced multicore Cell/B.E. processors
byteLAKE’s CTO

• Founded in 2016 by a dedicated group of business and technology professionals, we boast a
long history of collaboration that predates our official establishment.
• Our team members bring a wealth of experience, having worked with Fortune 500 companies
across Europe, the USA, and Asia. Furthermore, many of our team members have pursued
academic paths, earning DSc and PhD degrees in computer science, mathematics, and AI
applications.
• What truly sets us apart is our ability to blend the worlds of business and academia seamlessly.
This unique fusion empowers us to drive innovation in unprecedented ways, translating
theoretical knowledge into practical, real-world solutions.
• byteLAKE is home to an exceptional team of AI and HPC specialists and researchers,
handpicked for their expertise and unwavering commitment to continuous learning. This
combination of talent and dedication places us at the forefront of the AI landscape, well-
equipped to tackle the most intricate challenges with confidence.
byteLAKE Team

AI Solutions for Industries

Recommended

Recommended

More Related Content

Similar to AI Solutions for Industries

Similar to AI Solutions for Industries (20)

More from byteLAKE

More from byteLAKE (20)

Recently uploaded

Recently uploaded (20)

AI Solutions for Industries