1. H2O.ai Confidential
Slide 1 - Intelligent Experiences
There is an undeniable shift in the world towards intelligent Manufacturing
This creates big stakes and huge urgency
Slide 2 - History Leading up to this Shift
Slide 3 - There will be winners and losers (maybe show tech giant market cap, or compare top 10 companies from 2000 to today.
Slide 4-5 - Winner examples (tech giant, start-up, incumbent) ///missing in this deck
Slide 6 - Relevance of AI in the Customer’s Industry
X% of <insert industry> executives think that…
Here are the areas / use cases where AI is having real impact in your industry (reference would be in this slide)
Slide 7 - Promised Land (vision of their industry based on AI ubiquity) //missing
Slide 8 - Point-of-View
Here are some specific use cases that align to your organizational priorities, plus
Potential value impact they could drive over the next 3 years (e.g. value hypothesis)
Slide 9 - H2O.ai
AI Cloud Platform - Make. Operate. Innovate.
Old World vs. New World set up (just like the video)
Introduce features as “magic gifts”
Slide 10 - Case Study
Here’s an example where we’ve worked with one of your peers, plus
The value they achieved (e.g. use case benefit + H2O benefit)
Present evidence that you can make the story come true
Section 2: Transformation and Close
Slide 11 - Maturity Levels of AI Transformation
How the customer should approach this transformation organizationally
The major phases they’ll go through as they move towards maturity
Key considerations to drive success in each phase
Slide 12 - The Path to Success
Use case discovery + business case development
AI strategy + technical evaluation
Contracting
Deployment + development of use cases
Create/augment applications and business processes
Realize and track value
Slide 13 - Next Steps
Use Case Discovery Workshop (w/LOB stakeholders)
AI Strategy Workshop (w/technical stakeholders)
Manufacturing AI Transformation Narrative
(Audience - Transformational Executive / Business Exec)
2. H2O.ai Confidential
H2O.ai - AI for
Manufacturing
Improve yield and throughput while
reducing scrap, inventory, and cycle times
with the H2O AI Cloud
Sean Lee
Director, powergen.ai
3. H2O.ai Confidential
v
Intelligent Experiences Have Come to Manufacturing
Intelligent
Supply Chain
Recommended Raw Material
and Component Orders
Sales Predictions for
Finished Product (Demand
Sensing)
Return Forecasting
Predictive
Maintenance
Predict Equipment Failure
Optimize uptime vs.
maintenance cost
Predictive Failures
Predict failures down the
line for current work-in-
process product
Identify failures using video
and images
Intelligent prediction of
product that will fail in the
field, driving to zero defect
4. H2O.ai Confidential
v
Digital, Data & AI Have Powered This Transformation
1980 2000 2015
1990 2010 2020
Computers
The availability of personal
computers begins the
shift towards digitalization
Digital
Experiences
The proliferation of smart
devices and mobility
shifted B2C and B2B
engagement patterns
towards digital
AI
The rise of cognitive
computing techniques
enabled machines to
gather insights and make
decisions
Internet
The commoditization of
the Internet began
moving the world online
Big Data
The growth in digital
footprints and connected
devices created an
explosion of data that
could be gathered and
analyzed
Intelligent
Experiences
Automated decision
making by machines
created an opportunity to
deliver intelligent,
experiences at scale
5. H2O.ai Confidential
The Age of AI is Now
84%
85%
of executives say they
won’t achieve growth
objectives without
scaling AI.
75%
71%
of executives believe
they risk going out of
business in 5 years if
they don’t scale AI.
76%
76%
of executives
acknowledge they know
how to pilot, but struggle
to scale AI across the
business.
Global United States
Reference: Scaling AI: From Experimental to Exponential (Accenture)
7. H2O.ai Confidential
[Flex Example] Value Summary
Accelerating AI Transformation at Flex
Over the past three months Flex has evaluated H2O’s AI platform for the development and
operation of machine learning applications that directly support their strategic priorities for
Manufacturing Excellence, Technology and Domain Expertise and Global Scale and Reach.
As Flex seeks to expand gross margins and optimize product mix regionally, these applications
are designed to drive process efficiencies, yield improvements and plant capacity increases
while simultaneously optimizing the infrastructure and personnel costs associated with
integrating AI into the daily operations of the business.
Through the course of the evaluation, it was demonstrated that use cases in Real-Time
Anomaly Detection and Test Time Reduction could drive yield improvements of up to 4% as
well as test time reductions of up to 40%. Moreover, it was proven that H2O ”democratizes
AI” by enabling no-code or low-code Flex users to create highly accurate predictive models,
deploy production-grade applications quickly, and more easily maintain those applications to
ensure a high degree of trust and confidence amongst business users.
H2O Aligns to Flex’s Strategic Priorities
Enable a global footprint to regionalize at scale
Focus on Mix to increase higher value business
Leverage industry-leading advanced capabilities to achieve manufacturing
excellence
Continue to innovate in manufacturing through Industry 4.0, full lifecycle
management and capabilities for regulated industries
Utilize cross-business synergies to enable a rapid exchange of technology and
best practices
Leverage technology and domain expertise to create operational efficiency
through IoT, human-machine interface and automation
Create resilient supply chains built on regional solutions
Navigate global complexities through a digitized supply chain, a resilient global
network and real-time visibility and analytics
1
2
3
Enterprise Multi-Site Investment Strategy for Flex
$27.2M
TOTAL INVESTMENT
19.5x
ROI
120 DAYS
PAYBACK PERIOD
3-Year Value Potential for Flex
COGS Savings
COGS SAVINGS ($ MILLIONS) – 2022 – 2025e
$532M Profitability
EBIT MARGIN (%) – 2022 – 2025e
+1.2pp
$28M
$154M
$350M
+0.1pp
+0.5pp
+1.2pp
Theme H2O AI Use Cases
Horizo
n
Plants Investment
Intelligent
Manufacturing
Real-Time Anomaly Detection
Apply computer vision to assembly line
images to detect anomalies in the
assembly process that lead to IST and
SLT failures and raise alerts/feedback in
real time
1 – FY23 10 $3.8M
2 – FY24 25 $7.8M
Test Time Reduction
Apply machine learning to IST and SLT
test logs to extract process
optimization insights for early failure
detection and overall test time
reduction
3 – FY25 50 $15.6M
4
9. Democratize AI with H2O.ai
200K
Community & Companies
25
Kaggle Grandmasters
World’s #1, #2, #5, and #9
8
7
222OF
THE
H2O
OF THE TOP 10
BANKS
OF THE TOP 10
4 OF THE TOP 10
MANUFACTURING
COMPANIES
INSURANCE
COMPANIES
Founded in Silicon Valley, 2012
Investors (Series E): Goldman Sachs, Ping An, Wells Fargo,
NVIDIA, Capital One, Nexus Ventures,
Commonwealth Bank Australia
FORTUNE
500
10. H2O.ai Confidential
Long AI Projects that Add No Value
Data Scientist Has an
Idea or is Told to work on
a Project
Step 3
Works with custom-built systems
or can’t get the model in
production
Toil on a notebook creating
features and optimizing to build
a “perfect” model
Step 2
Step 1
Step 4
Hands predictions to the business user,
who doesn’t understand the model or why
the work was done in the first place
Step 5
Not Used
0 9
MONTHS
11. H2O.ai Confidential
0 2
WEEKS
Fast And Successful AI Projects
Data Scientist, Business Analyst, Developer,
or Data Engineer Works with Business
Owner on the Problem
Explains the model to the business owner and
iterates with the business owner to ensure
simplicity and success once it’s in production
Uses No Code or AutoML services and builds
highly accurate models in days or hours
1-Click to production, and simple registration to
provide MLOps with a single pane of glass for
every model in the organization
Low-code departmental app
is built, or AI is integrated
into existing apps or
databases. It is used by
business owners, and
delivers a high ROI
Step 3
Step 2
Step 1
Step 4
Step 5
13. H2O.ai Confidential
Make. Operate. Innovate.
Data Intelligence
Feature Engineering
Feature Store
Auto ML
Forecasting
NLP
Computer Vision
Documents
NEW Deep Learning
White Box Models
Interpretability Methods
Bias Detection
Third Party Models
UI Creation
ML Integration
Prototype Dev
Model Repository
Model Deployment
Model Monitoring
Data Science Apps
Vertical Apps
Horizontal Apps
Flexible Architecture
Extensible
Distributed
Scalable
Feature
Transformation
Machine
Learning
Explainable
AI
Low Code
App Development
Machine Learning
Operations
AI
AppStore
Purpose
Impact
Democratize AI
Data
Idea
H2O AI CLOUD
Democratize and Accelerate AI Results with Trust and Confidence
14. H2O.ai Confidential
AI Platform Requirements and
the H2O AI Cloud
H2O AI Cloud is the most comprehensive AI
Cloud, that meets the demanding
requirements of the largest enterprises and
the fastest growing startups.
AI Clouds deliver faster time-to-value,
optimize AI’s impact, and make it much
simpler and lower risk to manage and govern
AI across an organization.
H2O AI Cloud has a set of products and
capabilities, that help it stand out against
every key AI Cloud platform requirement.
H2O AI Cloud
Support all AI Use Cases
● Big Data >1TB
● Structured Data
● Time-Series Data
● Text, Image and Audio Data
● Document Data
Deliver Rapid Time-to-Value
● Advanced AutoML
● No Code Deep Learning and Document AI
● 1-Click Deployment to Production
● Apps for Business Explainability
● Low Code AI App Development
● AI Apps for Users
Support Multiple Users /
Democratize AI
● Code Interfaces
● No Code Interfaces for all Data Types
● #1 No Code AutoML
● Snowflake SQL Interface Integration
● AI AppStore
Easy to Explain, Monitor, and
Govern AI
● Most Robust Explainable AI Capabilities
● Model Registry with Hosting, Scoring, and Monitoring
● Single Pane of Glass for models in DBs, apps, H2O MLOps,
existing tools
Provide the Highest Accuracy
● AutoML produces the highest accuracy results with the fewest
iterations
● 100s of Kaggle Grandmaster Optimized Recipes for Use cases
● Deep Learning and Document AI Optimized by Kaggle GMs
Integrate into Existing Data,
Apps and AI Tools
● No infrastructure management, eliminates undifferentiated
heavy lifting
● Any Cloud, Multi-Cloud, or Hybrid Environment
● Integrates with all existing data, AI tools and apps
15. v
H2O.ai Confidential
Fastest Time-to-Value for
any Use Case
#1 AutoML
#1 Document AI
Interfaces for Multiple User Types
No Code Deep Learning
250+ Optimized Recipes
100+ Pre-build AI Apps
1-Click Deployment
AI AppStore to Collaborate with the
Business
Most Comprehensive
Explainable AI
Multiple Whitebox Models
Multiple 3rd Party Models
The Most Interpretability Methods
Automatic Documentation
Bias Detection
Most Flexible Architecture
Cloud Agnostic
Run Models Anywhere
Fully Managed Offering
Integrate with Existing Tools,
Data, and Applications
Why H2O AI Cloud?
18. H2O.ai Confidential
Computer Vision Adaptive Manufacturing
Flex Use Case: Enhanced Quality Control with Video | 100 Images per Product Scored in 35 Seconds
1
Make Model
2
Deploy Model
H2O AI Cloud
3
Score Model
H2O
eScorer
Multiple Remote
Factories
Camera at factories shoot video
for defect detection
Nvidia Server creates 100
.jpgs from video
Model Monitoring
with H2O MLOps
Predictions used for Drift
and Bias Analysis
19. H2O.ai Confidential
NXP Predicts Wafer Failure with H2O AI Cloud
Manufacturing
Leading Semiconductor Manufacturer
The auto industry is moving to zero
defect, and we’ve previously never
been able to predict failure at low
PPM and PPB levels. The feature
engineering and insights of H2O
Driverless AI helping us solve this very
challenging problem.”
Jim Bird, Head of Data Science for
Manufacturing
● Silicon wafers are tested before they are
cut to become chips.
● 1000+ QA electrical tests, but nothing
multi-variant.
● Some chips fail in the field, and auto
industry moving to zero defect.
● Working with H2O AI Cloud to predict
wafer failures without yield impact.
Challenges Why H2O AI Cloud?
● Automated Feature Engineering Insights
with H2O Driverless AI.
● Model documentation and insights from
the 1000s of models that H2O Driverless
AI builds.
20. H2O.ai Confidential
1. Predict Equipment Failure
2. Predict the Remaining useful life (RUL)
3. Early warning system for all the systems
at an oil rig
Oil & Gas Equipment Failure
Use Case
Customer: Petroliam Nasional Berhad (PETRONAS) is the custodian of Malaysia’s national oil and gas resources,
they explore, produce and deliver energy to meet society’s growing needs.
1. Faster and superior Model training and prediction
performance on GPU and Java Scoring Pipeline for
Nano second inferencing.
2. 17K machine sensor data coming in every min and
historical downtime data in their SAP system.
3. Huge Data (Machine Data), high frequency problem
THE RIGHT INFORMATION IN THE RIGHT TIME” TO PREVENT “UNPLANNED STOPS”
Business Objective Challenge
Rare Anomalies Time Series Problem
1. An end to end pipeline using HO DriverlessAI to process the data, train a machine learning model, and
generate inference.
2. Driverless AI is deployed on a 2 X NVIDIA® DGX-2™ server with 16 GPUs only for the Peninsula for a start
and plan to expand across Malaysia
3. Managed to predict the machine failure before 3 days
Enterprise AI Solution with H2O.ai
Data: High Frequency Big Data Rare Anomaly
Detection Problem
● 380 oil rigs
● 64 equipments per rig
● 17k sensors, data captured over 5 seconds,
● stored SAP systems
● Anomalies : 1 out of 10,000 instances
21. H2O.ai Confidential
Coke Prediction in Production
Use Case
In Platformer unit, Platinum (Metal), Aluminum (Acid) based catalyst (R264x, R364x) supplied by UOP is used for
converting paraffins & Napthenes to Aromatics (Reformate).
• A robust model to predict coke can help us in precision optimization of plant
parameters to mitigate coke formation & thereby avoiding spurious cut
down of plant throughput.
• However, methods tried to predict this coke (wt%) using 1st principle
models, simulation models haven’t been very accurate.
Model: Robust Coke prediction Model
Problem:
The reactions for Aromatics production result in coke (95% Carbon content) formation on the surface of
the catalysts.
This coke fills up the active sites on the catalyst making it ineffective for further reaction. Hence, this coke
needs to be removed (regeneration) continuously for continuous production of reformate.
Early Event Detection (EED) to take proactive actions to
control high coke and CCR upsets (reduction of CCR upsets on
high coke –events less than 2 per year)
01
02
Identify the factors which are causing high Coke - should
predict accurately 95% of time
22. H2O.ai Confidential
Control Valve Reliability
Use Case
Existing maintenance strategies provide assurance using frequency-based inspections like overhauling
based upon number of hours the valve is in service and hence are capable only of providing spot values
across the lifecycle of the valve.
With online data being available in valves using positioners, specific data points pertaining to CVs and
impacting process parameters can be evaluated.
Business Goals:
1. Improve uptime of Control Valves
2. Predict valve performance over specific
periods & conditions
3. Energy optimization by monitoring
leakages & recommending setpoints
4. Specify Control Valves Job scope for major
plant turnarounds
5. Reduce Maintenance cost of overhauls
AI Driven Solution
1. Increased reliability and uptime of CVs due
to Online health assessment and
monitoring
2. Optimizing process control by monitoring
Intelligent Valve Positioners (IVPs)
3. Energy optimization by precise monitoring
leakages and pneumatics
4. Cost reduction by institutionalizing
condition-based maintenance for CVs’
Subsystem 5:
Power supply; (electric, air,
hydraulic)
Subsystem 4:
DCS or control signal
Subsystem 1:
Control valve; I/P, positioner,
and actuator
Subsystem 2:
Process (Fluid)
Subsystem 3:
Design (sizing, material
selection)
Subsystem 6:
Environment
(Support,
accessibility, fire
protection)
23. H2O.ai Confidential
Business Background
Data quality is the key importance to performance of predictive maintenance. In upstream asset
Shearwater low pressure compressor has numerous missing values at various row positions.
Meanwhile, many tags have large periods of missing data and most time-consuming part of data
preparation is arguably spent trying to assess which tags to keep and impute or drop altogether.
Compressor Data Quality Assurance
Use Case
Use Case / Approach
● Customer used H2O.ai platform to successfully and accurately predict values to use for
imputation will greatly aid the time spent cleaning data and identifying good features.
● H2O.ai provided realistic predictions of missing values for a set of arbitrarily chosen tags
● H2O.ai also provide predictions for all regions of data that have been flagged as missing and
were benchmarked against a series of blind experimental tests
Shell.ai
Self- Service and Professional AI Platform in
collaboration with H2O.ai
24. H2O.ai Confidential
High
lifetime
Low
lifetime
Battery Life Prediction
Use Case
Battery Life Prediction using machine learning for LG Chem (part of LG Group)
Under various usage patterns, conditions, and data pattern is far from obvious, understanding
driving factor is very difficult. Customer used H2O auto ML to derive insights about life affecting
factors, they also created a model to predict (or estimate) the remaining life of batteries in
different machines.
Machine learning Battery life prediction
Your Battery Lifetime
25. H2O.ai Confidential
Energy Consumption Prediction
Use Case
PJM Interconnection LLC (PJM) is a regional transmission organization (RTO) in the United States.
It is part of the Eastern Interconnection grid operating an electric transmission system. They
wanted to predict the energy consumption by hour.
Business Problem:
Predict Hourly Energy Consumption by locality / area
Predicting how much energy an equipment (or machine) will consume by time
Train and Test Data Predictions / Forecasts over time
MAPE error in the range of 7%
Impact:
Resource Optimization, Identifying worse areas, identify failures
26. H2O.ai Confidential
Demand Sensing in Supply Chains
Use Case
Solution:
Time Series forecast provides predictions for the sales next X number of weeks and can
run every week. The current crisis will almost always affect the forthcoming sales.
Forecasts #s can be updated to produce a second prediction/correction of the sales of the
Y number of weeks and can be run daily/hourly. These daily/hourly data could be new
COVID cases, POS sales, social media data (tweets) etc.
Demand Sensing can alter existing short-term time series predictions and adjust for day to day, hour to hour or even near real-time
History
A Wave App that Forecasts Sales with Augmented Data:
Wave Demand Sensing for sales forecasting uses augmented COVID-19 data. It can help a company
forecast the sales for different SKUs filtered by region/customer/brands.
The forecasting model uses sensing variables such as –COVID-19 cases, social sentiment and more.
The Demand Sensing app then shows the impact on sales with or without sensing variables.
Times Series modelling
Week -4 Week -3 Week -2 Week -1 Week 0 Week 1 Week 2
History Present Future
Demand Sensing modelling
Future
Days
Week 1 Week 2
…
…
Short Term Time Series Forecasts
27. H2O.ai Confidential
H2O.ai Confidential
Jump Server
Jump Server needed for
“Offline” Factory to
Snowflake
Model (MOJO)
F100 Customer - Adaptive Manufacturing with Multiple Offline
and Online Manufacturing Facilities
Input and Predictions Outputs, Including Explainability at Scoring
are Returned to Snowflake for analysis
Use Case: Enhanced Quality Control with Predictive Failure
1
Data to Train Model
H2O Scoring
H2O Scoring
2
Make Model
3
Distribute Model
through Snowflake
4
Score Model at Factory
5
Model Registry and
Monitoring with
H2O MLOps
29. H2O.ai Confidential
H2O.ai Confidential
DEFINE
Create AI objectives
THE PHASES OF
AI MATURITY
PILOT
Select and deliver
projects that prove
value
SCALE
Develop a framework
for scale across the
business
ENABLE
AI is built into the
fabric of day-to-day
problem solving
EVOLVE
Monetizing AI
through intelligent
products
80%
of organizations are stuck in the
“proof-of-concept factory”
Lack of supportive
organizational structure
Absence of foundational
data capabilities
Lack of business adoption
Source: Accenture 2021
30. H2O.ai Confidential
H2O.ai Partners with Customers to Achieve Successful AI
Transformations
AI Value Creation Journey
Value
Creation
Decision Support &
Augmentation
Decision
Automation
PILOT SCALE ENABLE EVOLVE
AI Business
Implementation
Make. Operate. Innovate.
Feature
Transformation
Machine
Learning
Explainable
AI
Low Code
App Development
Machine
Learning
Operations
AI
AppStore
AI Technology
Product Creation