Weitere ähnliche Inhalte Ähnlich wie Data Natives meets DataRobot | "Build and deploy an anti-money laundering model in 20 minutes" - Kayne Putman & Christian de Chenu (20) Mehr von Dataconomy Media (20) Kürzlich hochgeladen (20) Data Natives meets DataRobot | "Build and deploy an anti-money laundering model in 20 minutes" - Kayne Putman & Christian de Chenu1. Confidential. ©2019 DataRobot, Inc. – All rights reserved
Kayne Putman
Customer Facing Data Scientist
DataRobot
Build and deploy an
Anti-Money Laundering
model in 20 minutes.
Christian de Chenu
Customer Facing Data Scientist
DataRobot
2. Confidential | Copyright © 2018 | DataRobot, Inc.
Strawpoll
● Who has heard of (or is working with) DataRobot
● Who is working on an AML use case in their organisation?
3. Confidential. ©2019 DataRobot, Inc. – All rights reserved
1. Introduction to DataRobot (3 mins)
2. Introduction to AML (5 mins)
3. AML demo (15 mins)
4. Real AML use case (5 mins)
5. Q&A
Agenda
4. Confidential. ©2019 DataRobot, Inc. – All rights reserved
1. Introduction to DataRobot
2. Introduction to AML
3. AML demo
4. Real AML use case
5. Q&A
5. Confidential. ©2019 DataRobot, Inc. – All rights reserved
2012
Invented Automated
Machine Learning
$224 M
In funding to date
1.2B+/2.5M daily
Models built on DataRobot in the cloud
800+
Worldwide employees
4 Major;
40+ Minor
Product releases per year
Thousands
Successful AI Projects
The World’s Most Trusted Automated Machine Learning Platform
INSURANCE BANKING FINTECH HEALTHCARE TELECOM GOVERNMENT RETAIL MANUFACTURING MANY MORE
6. Automated Machine Learning: The benefits
2020 2022 2024
Increase Data
Scientist
Productivity
Empower
Data Analysts,
Engineers, and
other SMEs
2008 2010 2014 2016 2018
Companies are achieving AI success with
automated machine learning
AND the team they already have in place
2012
Demand for machine learning & AI
Supply of data scientists
7. Confidential | Copyright © DataRobot, Inc. | All Rights Reserved
ATTRIBUTES
1. Knowledge of the overall & specific missions
2. Knowledge of the data
3. Ability to write code to gather data
4. Ability to write code to explore/inspect data
5. Ability to write code to manipulate data
6. Ability to write code to extract actionable intel
7. Ability to write code to build models
8. Ability to write code to implement models
9. Foundational statistics
10. Internals of algorithms
11. Practical knowledge and experience
12. Knowing how to interpret and explain models
DATAROBOT DEMOCRATIZES DATA SCIENCE
Domain
Expertise
Math &
Stats
Domain
Expertise
Programming
Skills
9. Confidential | Copyright © 2018 | DataRobot, Inc.
Building an AI-Driven Culture
“AI will crush you!”
OK, we’ll try a POC.
Wow! $20M ROI from one
use case! What else can we
do?
Great! Our Automation-first approach is
delivering much more value to the business, and
faster.
Let’s empower and mentor more data-savvy
talent.
We’re
changing the
game in our
industry.
Our groups are working
together on AI projects
driving better
outcomes.
AI-Driven Enterprise
Embed AI in all
business processes
Solution Acceleration
Solve urgent business
needs faster
Data Scientist Productivity
Get more projects
done in less time
Data Science Democratization
Empower your existing
team to build AI
10. Trusted AI Partner Every Step of the Way
DataRobot University
Provide practical, hands-on education for DataRobot
users, from executives to analysts and data scientists,
no matter what your data science experience may be
24/7 customer support
Support customers through specialist knowledge that
is equipped and empowered to fix any issue you may
encounter, and ensure you're set up for success
Account Executive
Understand your objectives
and recommend the best
DataRobot resources
Field Engineer
Integrate DataRobot with
customer infrastructure and
production pipelines
AI Success Manager
Assess and coordinate
resources needed to
achieve ongoing success
Customer Facing Data Scientist
Deliver data science consulting to
support use cases and enable value-
producing users
Account
Team
11. Confidential. ©2019 DataRobot, Inc. – All rights reserved
Technology Alliances
Cloud RPA Analytics
WorkbenchesSolutions
Data & Hardware
Platforms
Amazon SageMaker
12. Confidential. ©2019 DataRobot, Inc. – All rights reserved
1. Introduction to DataRobot
2. Introduction to AML
3. AML demo
4. Real AML use case
5. Q&A
13. Confidential | Copyright © 2018 | DataRobot, Inc.
Money Laundering
(noun)
Making funds obtained from illegal activity appear
legitimate by concealing the source
15. Confidential | Copyright © 2018 | DataRobot, Inc.
• Anti-Money Laundering (AML) Compliance Program pillars:
○ Know Your Customer (KYC) to establish customer identity and risk
○ Transaction Monitoring to detect potential money laundering activity
→ Suspicious Activity Reports (SAR) filed with the National Crime Agency in the UK.
• Some recent penalties from the FCA (Financial Conduct Authority) for AML
non-compliance:
March 2019
£27.6 million
March 2019
£34.3 million
April 2019
£102.2 million
How is Money Laundering prevented?
16. Confidential | Copyright © 2018 | DataRobot, Inc.
Today, rule-based systems are used to refer potentially suspicious activity for manual
review by internal investigators, which can result in a SAR.
Transaction Data:
Deposits, Payments, ...
Transaction
Monitoring
Rules
(can be billions of rows)
No Alert
Alert:
Potentially
Suspicious
Expert
Manual
Review
Not
Suspicious
Suspicious
File
SAR
High False
Positive Rate
BEFORE: Rule-based transactional monitoring
17. Confidential | Copyright © 2018 | DataRobot, Inc.
Alert:
Potentially
Suspicious
Expert
Manual
Review
Customer
Order of
Review SAR Filed
Bob 1 No
Barbara 2 No
Walter 3 No
Bill 4 No
Bonny 5 No
Wilma 6 No
Barry 7 No
Bart 8 No
Brittany 9 No
Willy 10 Yes
Number of
Manual Reviews
False Positive Rate
SARs Filed
10
1
90%
BEFORE: Results
19. Confidential | Copyright © 2018 | DataRobot, Inc.
False positive rates can be reduced using automated machine learning to yield a set of
ranked alerts with a higher concentration of SARs.
Transaction
Monitoring
Rules
No Alert
Alert:
Potentially
Suspicious
Not
Suspicious
Suspicious
File
SAR
Reduced False
Positive Rate
*NEW
Full
Manual Review
of higher
quality Alerts
AFTER: Monitoring with Automated Machine Learning
20. Confidential | Copyright © 2018 | DataRobot, Inc.
Alert:
Potentially
Suspicious
Number of Full
Manual Reviews
False Positive Rate
SARs Filed
1 4
90% 60%
Full
Manual Review
of ranked, higher
quality Alerts
Time to find 10 SARs would drop from 30 hours to ~7.5 hours
10Customer
Order of
Review SAR Filed
Bob 1 No
Barbara 2 No
Walter 3 No
Bill 4 No
Bonny 5 No
Wilma 6 No
Barry 7 Yes
Bart 8 Yes
Brittany 9 Yes
Willy 10 Yes
AFTER: Results
21. Confidential. ©2019 DataRobot, Inc. – All rights reserved
1. Introduction to DataRobot
2. Introduction to AML
3. AML demo
4. Real AML use case
5. Q&A
22. Confidential. ©2019 DataRobot, Inc. – All rights reserved
1. Introduction to DataRobot
2. Introduction to AML
3. AML demo
4. Real AML use case
5. Q&A
26. Confidential | Copyright © 2018 | DataRobot, Inc.
Thought-provoker:
In an AML use case, which
feature(s) might be the most
impactful?
27. Confidential | Copyright © 2018 | DataRobot, Inc.Source: http://www.scielo.org.za/scielo.php?script=sci_arttext&pid=S2224-78902017000200002