Today, organizations deploy more AI/ML workloads on AWS than on any other cloud platform. The cloud has removed many of the challenges associated with scalability, and it’s never been easier or more cost effective to build custom and intelligent data models. In this session, learn how the C3 Platform leverages the full power of Intel Xeon Scalable processors on AWS to rapidly train, deploy, and operationalize AI/ML and big data applications like C3 Inventory Optimization and C3 Predictive Maintenance. In addition, a customer shares how these solutions helped achieve demonstrable value. This session is brought to you by AWS partner, Intel.

2. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Accelerate AI/ML adoption with Intel®
processors and C3 Platform on AWS
S e s s i o n I D
Binay Ackalloor
Dir of Bus Dev, AIPG
Intel
Dib Banerjee
Director of Products
C3

3. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved.
Legal Notices & Disclaimers
This document contains information on products, services and/or processes in development. All information provided here is
subject to change without notice. Contact your Intel representative to obtain the latest forecast, schedule, specifications and
roadmaps.
Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service
activation. Learn more at intel.com, or from the OEM or retailer. No computer system can be absolutely secure.
Tests document performance of components on a particular test, in specific systems. Differences in hardware, software, or
configuration will affect actual performance. Consult other sources of information to evaluate performance as you consider your
purchase. For more complete information about performance and benchmark results, visit http://www.intel.com/performance.
Cost reduction scenarios described are intended as examples of how a given Intel-based product, in the specified circumstances
and configurations, may affect future costs and provide cost savings. Circumstances will vary. Intel does not guarantee any
costs or cost reduction.
Statements in this document that refer to Intel’s plans and expectations for the quarter, the year, and the future, are forward-
looking statements that involve a number of risks and uncertainties. A detailed discussion of the factors that could affect Intel’s
results and plans is included in Intel’s SEC filings, including the annual report on Form 10-K.
The products described may contain design defects or errors known as errata which may cause the product to deviate from
published specifications. Current characterized errata are available on request.
No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.
Intel does not control or audit third-party benchmark data or the web sites referenced in this document. You should visit the
referenced web site and confirm whether referenced data are accurate.
Intel, the Intel logo, Pentium, Celeron, Atom, Core, Xeon, Movidius and others are trademarks of Intel Corporation in the U.S.
and/or other countries.
*Other names and brands may be claimed as the property of others.
© 2018 Intel Corporation.

4. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Agenda
AWS + Intel – A Decade of
Collaboration
Intel AI
C3 + Intel + AWS

5. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AWS + Intel
A decade+ of collaboration
10+ year engineering partnership
Digital Transformation
Shared customer passion
High performance + low costs
World class supply chain
Cloud &
Data Center
Things &
Devices
AWS IoT and
Dev Kits
Alexa
Voice Services
Amazon EC2 Amazon EBS
AI/ML
HPC/Analytics
Big Data/SAP on AWS
Hybrid Cloud/VMware on AWS
IoT/Edge Computing
COMMON HISTORY & VALUES JOINT PRIORITIES

6. © 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Confidential and Trademark© 2018, Amazon Web Services, Inc. or its Affiliates. All rights reserved. Amazon Trademark
M L F R A M E W O R K S &
I N F R A S T R U C T U R E
A I S E R V I C E S
R E K O G N I T I O N
I M A G E
P O L L Y T R A N S C R I B E T R A N S L A T E C O M P R E H E N D L E XR E K O G N I T I O N
V I D E O
Vision Speech Language Chatbots &
Contact Centers
M L S E R V I C E S
A M A Z O N
S A G E M A K E R
F P G A s
Frameworks Interfaces
Infrastructure
IoT EC2 CPUs Edge EC2 GPUs
AWS
DEEPLENS

7. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Bringing your AI Vision to Life

8. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Intel AI Hardware
Multi-purpose to purpose-built from Device to Cloud

9. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Intel AI Portfolio*
Portfolio of software tools + hardware
to accelerate time-to-solution

10. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
AI customer example
Intel works with customers across the entire AI lifecycle
TIME-TO-
SOLUTION
Opportunity Hypotheses Data Modeling Deployment Iteration Evaluation
15% 15% 23%
15% 15%
8% 8%
Experiment with
Topologies
Tune Hyper-
parameters
Share
ResultsLabel Data Load Data Augment Data
Support
Inference
Compute-intensiveLabor-intensive Labor-intensive
Proof of
concept
Training
Source Data Scale & Deploy Inference Scale & Deploy inference within broader application15%
15%
23%
15%
15%
8%
8%
Dev Cycle
…Build, Deploy
& Scale

11. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
EDGE/DataCenter
Source: Intel customer engagement
*Other names and brands may be claimed as the property of others
Inference
Label Store
Media
Server
Inference
Media
Store
Training Model Store
Training
Feature
Engineering
Data Ingest
Analytics/
Solution
Layer
Service
Layer
Intel helps customers deploy & scale real AI solutions
ENDPoint
Inference 10 x
Data Ingestion
Inference
Tagged Datasets
Service Layer
Media Server
Data Ingestion
Data Ingestion
Data Ingestion
Inference
Inference
Inference
Tagged Datasets
Service Layer
Service Layer
Media Server
Media Server
Multi-Purpose Cluster:
4 nodes
One ingestion
per day, one-
day retention
Media
Server:
Storage
Media Store
Media Store
Media Store
Media Store
Media Store
Media Store
Training:
Model Store
Model Store
Model Store
Model Store
Label Store
Label Store
Label Store
Label Store
110 Nodes
8 TB/day per
camera
10 cameras
3x replication
1-year video
retention
4 mgmt nodes
4 nodes
20M frames
per day
2 nodes
Infrequent op
3 nodes
Simultaneous
users
3 nodes
10k clips
stored
16 nodes <10 hours TTT
4 nodes
1-year of
history
4 nodes
Labels for
20M frames
/day
Data
Storage:
Per Node
1x 2S 61xx
20x 4TB
SSD
Training
TrainingPer Node
1x 2S 61xx
20x 4TB
SSD
Per Node
1x 2S 81xx
5x 4TB SSD
Per Node
1x 2S 81xx
1x 4TB SSD
Drone:
Inference
Inference
10 Drones
Per Drone
1x Intel® Core™
processor
1x Intel® Movidius™
VPU
Real-time object
detection
Software: OpenVino™ Intel® Movidius™ SDK TensorFlow* Intel® MKL-
DNN
Drone
AI customer example

12. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Intel AI Builders Ecosystem 100+ Partners*
Builders.intel.com/ai

13. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Faster Performance, Lower TCO Using Amazon
EC2 C5 instances
13
Instance Types:
Broadwell-based: c3.4xlarge instance (16 vCPUs, 30GB memory)
Skylake-based: c5.4xlarge instance (16 vCPUs, 32GB memory)
Workload: C3 Platform time series normalization
Results: 27% performance improvement and 41% TCO benefit using c5 vs c3 instances
$84
$50
$0
$20
$40
$60
$80
$100
Series1 Series2
TCO Benefit
100
73
0
20
40
60
80
100
120
Series1 Series2
TotalCPUTime(hours)
Performance Improvement

14. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.

15. © C3 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
AI-based Supply Chain Applications Can Significantly Improve
Supply Chain Efficiency
1515
Warranty
Optimization
Manufacturing
Predictive
Maintenance
Profitable BOM Defect
Detection and
Prevention
Inventory
Optimization Lowest Cost
Provider
Supply
Network
Optimization
Aftermarket
Insights
Inventory and SupplyManufacturing
Telematics
Predictive
Maintenance
Telematics
Operations
Optimization
FieldCustomer
Pricing and
Quoting
Optimization
Next-
generation
CRM
Demand
Forecasting
& Stocking
Order to
Promise
Sales & Operations Planning
© C3 IOT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Warehouse management
GPS tracker
Fleet management Syndicated market data
Transaction history
Weather
Labor management
Mobile data
Offers
Inventory
Bill of Materials
Production schedule RFID
Vendor Management
Orders
Quality management Demographics

16. © C3 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Current Supply Chain Challenges
Lack of real-time, scalable, comprehensive predictive analytic tools lead to poor visibility and controls
16
Manual
Processes
Slow Analysis Unreliable Results
Fragmented Toolkit Inability to Scale Lack of Real Time Data
High Inventory Cost Low Service-level
Lack of End-to-end Supply
Chain Visibility
Executives
Analysts

17. © C3 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
C3 Inventory Optimization Value Proposition
17
For a company
with $1B in inventory
Reduction in Inventory
Holding Cost
30%+
Service Level
99%+
Accuracy in Predicting
Supplier Delays
80%+
AI/Machine
Learning Algorithm
Real-time
Scalability Integration
$300M+
Inventory Reduction
Additionally,
$100M+
savings in logistics
cost annually
© C3 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE

18. © C3 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
The C3 Platform is Designed to Process High-volume, High-
frequency, Disparate Data at Massive Scale
18
Additional data sources: Easy to include additional data
sources that may be good predictors of demand-side or supply-
side uncertainties, e.g. weather, supply network delays etc.
Integration
Real-time
Real-Time: Generates real-time recommendations as new data
comes in using continuous analytic processing framework on C3
Platform (as against generating recommendations once a quarter
or once a year)
Scalability
Scalability: Uses C3 IoT Platform to scale massively at Item-
Location level on cloud infrastructure (as against using on-
premise solutions with limited scalability)
Data are siloed in multiple systems
Data from various silos are often accessible only by tediously
collating information from other teams
Data and alerts are updated infrequently
Data updated on a monthly basis
Access to only recent history
Only months, not years, of data are available to alert
rules and analysts
Machine learning/AI: Uses machine learning for predicting
demand-side and supply-side uncertainties (as against static
formulas from textbook)
Data is constrained by simplistic rules
Static rules lack sophistication to full represent the
wealth of information in the raw data
Algorithm
Typical Legacy Inventory Optimization solutions

19. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Optimize Inventory Levels
for a $30B Global Discrete
Manufacturer
CASE STUDY
G L O B A L
D I S C R E T E
M A N U FA C T U R E R
19 © C3 IOT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE

20. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE20
Employees
Discrete manufacturer with broad
range of industrial equipment
Years in operation
Annual revenue$30B
60,000
180
G L O B A L
D I S C R E T E
M A N U F A C T U R E R
© C3 IOT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
$

21. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
G L O B A L
D I S C R E T E
M A N U F A C T U R E R
21
Annual economic value
28-52%
$100-$200M
Savings in inventory holding costs
Scaling application to 40+ global factories
across over 1000 product lines
Weeks Project
Completion
12 42
Files Rows of Data
9M 3.2M
Material Movement
Events
Build application to optimize inventory levels of one product line
with over 40k unique parts
Highly customized, made-to-order product with a complex bill of materials
© C3 IOT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
CHALLENGE
RESULTS
PROJECT
DETAILS

22. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Implementation Timeline
22
WEEK 0 WEEK 1 WEEK 2 WEEK 3 WEEK 4 WEEK 5 WEEK 6 WEEK 7 WEEK 8 WEEK 9 WEEK 10 WEEK 11 WEEK 12 WEEK 14
Data workshop
Data Discovery
UTA*
Week 1 starts
Mid Trial Review
(On-Site at C3)
Delivered required
data - first pass
Project Kickoff and data review
Data loaded
first pass
Delivered
finalized data
Start of week 6
Complete
data loading
Design Data Model, Analytics, Application & UI
Load Data
CONFIGURE –Tune MPR, Improve Algorithm, Build Application & UI
Optimization algorithm integration
with platform data
Complete analytics &
machine learning
Complete UI
configuration
Executive
Evaluation
Trial Demo
Discuss with Analysts, Prepare Next Steps
CUSTOMER MILESTONE
C3 MILESTONE
JOINT MILESTONE
Review Review

23. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
C3 Inventory Optimization
23
Material Movement Consumption
Lead Times
Reorder Parameters
Production Orders
Demand Forecast
Purchase Orders
Actual Inventory
Material Movement– Arrivals
C3 Inventory Optimization
Application
Jupyter iPython Notebook
integrated with the C3 Platform
C3 Ex Machina C3 Intelligence
C 3 T Y P E S Y S T E M
C3DataIntegratorTM
AI&VisualizationTools
A I - M A C H I N E L E A R N I N G
U I S E R V I C E S
S T R E A M S E R V I C E S
B A T C H S E R V I C E S
I N T E G R A T I O N S E R V I C E S
D A T A S E R V I C E S
S E C U R I T Y S E R V I C E S
HDFS Relational Distributed
Key Value
Metadata

24. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
We are able to Identify Parts with Potential Inventory Savings
24
Example: Part A
Large difference between Actual Inventory and Consumption
Number
Static Analysis: Analyze daily historical part inventories relative to actual part usage and re-order parameters to
identify opportunities to reduce (or increase) safety stock levels
Large
difference
between
inventory levels
and part
consumption
levels –
indicating
opportunity to
reduce
inventory

25. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Approach: Replay History Day by Day and Dynamically Identify
Best Safety Stock
25

26. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Next Day: Replaying History Day by Day
26

27. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Results from Optimization Algorithm
We are able to optimize inventory levels while explicitly taking on some risk that parts may run out
27
Constraint: Ensure inventory level > 0, with 95% confidence
Example: Part number
Current value of Safety Stock: 27
Optimized Safety Stock: 0
Inventory levels can be reduced by lowering safety stock to even zero,
while being confident of not running out of parts 95% of time
Number
Number

28. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
Example Results: Distributions of Optimized Inventory Levels and
Daily Part Consumption are Now Much Closer Together
28
Example: Part: ‘A’
Current value of Safety Stock: 45
Optimized Safety Stock: 0
Distribution of consumption and optimized inventory are much closer together
Number of Parts
Number
of Days
Optimized Inventory
Number
of Days
Actual Inventory

29. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
C3 Inventory OptimizationTM on AWS
29

30. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
21% Performance Improvement, 42% TCO Reduction
Migrating from m4.16xlarge instances to m5.12xlarge
30
Instance Types:
Broadwell-based: m4.16xlarge instance (64 vCPUs, 256GB memory)
Skylake-based: m5.12xlarge instance (48 vCPUs, 192GB memory)
Workload: C3 Inventory OptimizationTM - Stochastic Supply Chain Optimization
Results: 21% performance improvement and 42% TCO reduction using m5 vs m4 instances
Number of Simulations
TotalCPUTime(sec)
0
50,000
100,000
150,000
200,000
1 2 3
Series1 Series2
$144
$83
$0
$20
$40
$60
$80
$100
$120
$140
$160
Series1 Series2
TCO Benefit for 1 million simulations

31. 858
426
0
200
400
600
800
1,000
Series1 Series2
31
Faster training on Amazon EC2 R5 instances
R5 was custom built for AWS and designed for the most demanding AI workloads
Instance Types:
Broadwell-based: r4.16xlarge instance (64 vCPUs, 488GB memory)
Skylake-based: r5.12xlarge instance (48 vCPUs, 384GB memory)
Workload: C3 Predictive Maintenance – U.S. Air Force
Results: 50% performance improvement and 49% TCO reduction using r5 vs r4 instances
TotalCPUTime(minutes)
$426
$218
$0
$100
$200
$300
$400
$500
Series1 Series2
TCO ReductionPerformance Improvement

32. © C3 IoT 2018 | CONFIDENTIAL - DO NOT COPY, REPURPOSE OR DISTRIBUTE
C3 at re:Invent – How to Learn More
Visit C3’s Connected Café at
the Sands Expo, #1026
C3 Session:
Faster, Better, Cheaper: AI Apps in One-Tenth
the Time and Cost
Tuesday, November 27
12:15 PM - 1:15 PM– Venetian, Level 4,
Marcello 4505
32

33. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.

34. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.
Related Sessions
Wednesday, November 28
A Smarter, Sustainable Future: How Duke Energy Improves Energy Efficiency and
the Customer Experience with the Alexa Connected Home
1:45 PM – 2:45 PM | Aria West Level 3, Juniper 4
Tuesday, November 27
Faster, Better, Cheaper: AI Apps in One-Tenth the Time and Cost
12:15 PM - 1:15 PM– Venetian, Level 4, Marcello 4505
Monday, November 26
Extend HPC Workloads to Amazon EC2 Instances with Intel
1:00 PM – 2:00 PM– Aria, Level 3, Juniper 1

35. Thank you!
© 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.

36. © 2018, Amazon Web Services, Inc. or its affiliates. All rights reserved.