The future of FinTech product using pervasive Machine Learning automation - André Balleyguier (DataRobot)
0 likes•199 views
Machine learning and automated decisions are reshaping businesses by automating processes, optimizing customer interactions, and efficiently measuring risk. The future of fintech relies on pervasive use of machine learning, but scaling ML applications is challenging due to the scarcity of data scientists and the complex ML process. Automated machine learning can address these issues by simplifying and accelerating the ML lifecycle, enabling a wider range of users to develop and deploy models at scale across all business functions.
![Confidential. ©2018 DataRobot, Inc. – All rights reserved
The value of AI today in the enterprise
Machine
Learning Deep
Learning
$ $ $ $ $ $
$ $ $ $ $ $
$ $ $ $ $ $
$ Automate $ Optimize $ Produce Actionable Insights
Key: [Boring Stuff] [Deep Learning] [Machine Learning]](https://image.slidesharecdn.com/shiftmoneyconference2018-andreballeyguier-190213131139/85/The-future-of-FinTech-product-using-pervasive-Machine-Learning-automation-Andre-Balleyguier-DataRobot-18-320.jpg)
The future of FinTech product using pervasive Machine Learning automation - André Balleyguier (DataRobot)
- 1. Shift Money Nov 26th 2018 The future of FinTech using pervasive Machine Learning automation André Balleyguier - Chief Data Scientist EMEA at DataRobot
- 2. Confidential. ©2018 DataRobot, Inc. – All rights reserved About me... + Chief Data Scientist EMEA London Previously...
- 3. Agenda Confidential. ©2018 DataRobot, Inc. – All rights reserved 1. A story about Transformation 2. Machine Learning in Financial services Applications of Machine Learning today in the FinTech space 3. The curse of ML: Scalability Main challenges to scale the use of Machine Learning 4. Automated Machine Learning in FinTech Leveraging highly automated and pervasive machine learning systems to optimise all business lines
- 4. Confidential. ©2018 DataRobot, Inc. – All rights reserved Machine Learning: A story about Transformation1)
- 5. Confidential. ©2018 DataRobot, Inc. – All rights reserved A story about my grandmother Germaine...
- 6. Confidential. ©2018 DataRobot, Inc. – All rights reserved
- 7. Confidential. ©2018 DataRobot, Inc. – All rights reserved Letters from friends to the family?
- 8. Confidential. ©2018 DataRobot, Inc. – All rights reserved Does this look familiar?
- 9. Confidential. ©2018 DataRobot, Inc. – All rights reserved Does this look familiar?
- 10. Confidential. ©2018 DataRobot, Inc. – All rights reserved So many ways to capture this information... ● Public data sources ● No response to other communication channels ● No activity for years…
- 11. Confidential. ©2018 DataRobot, Inc. – All rights reserved So many ways to capture this information... ● Public data sources ● No response to other communication channels ● No activity for years… Fuzzy matching of names to public data sources Predict customer propensity to respond Predict customer dormancy Machine Learning
- 12. Confidential. ©2018 DataRobot, Inc. – All rights reserved Machine Learning: A story about transformation Machine Learning and automated decisions based on data are reshaping the way businesses operate and interact with their customers. ● Automation of processes that usually require humain judgement: Fraud, Compliance, Cash management, Screening of candidates, … ● Optimisation of customer interactions based on profile ● Efficiently measure risk taken by various activities like lending, etc.
- 13. Confidential | Copyright © DataRobot, Inc. | All Rights Reserved “ Those who rule data will rule the entire world. ” M A S A Y O S H I S O N CEO | Softbank
- 14. Confidential. ©2018 DataRobot, Inc. – All rights reserved The impending AI divide AI and ML will generate $2.9 TRILLION in business value and recover 6.2 BILLION hours of worker productivity by 2021. - Gartner Predictions (Forbes) -
- 15. Confidential. ©2018 DataRobot, Inc. – All rights reserved The Digital champions ● Heavy focus on digitization of consumer ● Investing a lot in AI ● Data-driven and innovative culture
- 16. Confidential. ©2018 DataRobot, Inc. – All rights reserved Machine Learning in Financial Services2)
- 17. Confidential. ©2018 DataRobot, Inc. – All rights reserved Demystifying the buzzwords Artificial Intelligence Data Science Statistical Modeling Machine Learning Deep Learning Machine Learning: Learning from the past to predict the future AI: Systems able to perform tasks that ordinarily require human intelligence
- 18. Confidential. ©2018 DataRobot, Inc. – All rights reserved The value of AI today in the enterprise Machine Learning Deep Learning $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ $ Automate $ Optimize $ Produce Actionable Insights Key: [Boring Stuff] [Deep Learning] [Machine Learning]
- 19. Confidential. ©2018 DataRobot, Inc. – All rights reserved All aspects of financial services are impacted Upsell/Cross-sell Customer loyalty, retention CUSTOMER JOURNEY / 360 Pricing analytics Churn reduction OPERATIONS Loan review COMPLIANCE HUB Loss forecasting/ Stress Testing PRODUCT & SERVICE EFFICIENCIES Investment research and client targeting Portfolio analytics and robo-advising REVENUE GENERATION Prospecting Marketing RISK / FRAUD Delinquent asset resolution Credit and prepayment risk Model Governance and Validation KYC/AML Underwriting Succession planning Fraud reduction (ATM, card, merchant services)
- 20. Confidential. ©2018 DataRobot, Inc. – All rights reserved What are key use cases for... DIGITAL LENDING PLATFORMS Lending Risk scoring Application fraud detection Product recommendation Customer targeting PAYMENT PLATFORMS Transaction fraud Anti-Money Laundering / KYC Customer complaint resolution Crypto recommendation INSURTECH Underwriting automation Dynamic Pricing Fast-track Claims handling Claims Fraud or Litigation detection Renewal optimisation CHALLENGER BANKS New client targeting Anti-Money Laundering / KYC Customer Service Automation Product Cross-sell Credit Risk Transaction Fraud Investment / Robo-advisors
- 21. Confidential. ©2018 DataRobot, Inc. – All rights reserved Bottom Line: The 3 pillars of ML value 1. Customize Financial Products to clients and improve Customer experience, interest and retention 2. Provide access to services for Customers who usually don’t have access to them 3. Improve Operational Efficiency: Faster/Automated underwriting, Easier Compliance, More efficient Fraud detection, Optimised Customer Support
- 22. Confidential. ©2018 DataRobot, Inc. – All rights reserved Recent stories...
- 23. Confidential. ©2018 DataRobot, Inc. – All rights reserved The curse of ML: Scalability 3)
- 24. Confidential. ©2018 DataRobot, Inc. – All rights reserved Recent stories... “When it comes to data science at Monzo, we have a lot more ideas than we can practically implement. In particular, with machine learning there are many promising business applications but it’s prohibitively time consuming to test them all, especially as a small two-person team.” Dimitri Masin, Head of Data and Analytics
- 25. Confidential. ©2018 DataRobot, Inc. – All rights reserved The curse of ML “80% of Data Science projects never go to production!” Some (most) of my prospects “We mainly build prototypes” Scalability
- 26. Confidential. ©2018 DataRobot, Inc. – All rights reserved Why? Let’s go back to basics Discovery & Problem definition Prepare Data Develop model Socialise the model with the business Operationalise In theory, a Machine Learning project is a simple iterative flow:
- 27. Confidential. ©2018 DataRobot, Inc. – All rights reserved The Data Scientist: A rare unicorn 1. Knowledge of the business and business problem 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 items 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
- 28. Confidential. ©2018 DataRobot, Inc. – All rights reserved Machine Learning: So many things could go wrong! Discovery & Problem definition Prepare Data Develop model Socialise the model with the business Operationalise (1) This requires heavy business input, but a lot of data science teams work in silos (2) Data prep can be very time consuming: i.e put the data in the right shape for the algorithms (3) This requires a data scientist / engineer expert and can be hard
- 29. Confidential. ©2018 DataRobot, Inc. – All rights reserved Main bottlenecks in Machine Learning Demand for machine learning & AI Data scientists in the world 2010 2012 2014 2016 2018 2020 2022 20242008 ● Exponential demand in all sectors, including FinTech ● Experts are hard to find and retain ● Machine is a long and complicated process Our answer: Automated Machine Learning
- 30. Confidential. ©2018 DataRobot, Inc. – All rights reserved Automated Machine Learning for the Future of FinTech 4)
- 31. Confidential. ©2018 DataRobot, Inc. – All rights reserved Automated Machine Learning Open source technologies Years of experience of top-ranked data scientists Accelerate the process of researching, testing, and deploying Machine Learning models through automation and enabling a wider set of users. “Democratize” Data Science!
- 32. Confidential. ©2018 DataRobot, Inc. – All rights reserved Automated Machine Learning: more focused on the business expertise 1. Knowledge of the business and business problem 2. Knowledge of the data 3. Small amount of pragmatic education and mentoring.
- 33. Confidential. ©2018 DataRobot, Inc. – All rights reserved How does automation improve the process? Discovery & Problem definition Prepare Data Develop model Socialise the model with the business Operationalise (1) The process requires heavy business input, but a lot of data science teams work in silos => Democratise Data Science to business users through automation and education (2) Data prep can be very time consuming: i.e put the data in the right shape for the algorithms => Automation can automate some parts of the data preparation, also ensuring the process is more iterative, hence the data prep more focused on actual business value (3) Model development and deployment require a data scientist expert and can be hard => Automation allows to simplify the Machine Learning lifecycle and make it more accessible
- 34. Confidential. ©2018 DataRobot, Inc. – All rights reserved Automation addresses the skills challenge Demand for machine learning & AI Data scientists in the world 2010 2012 2014 2016 2018 2020 2022 20242008 ● Make Data Scientists more productive and closer to the business ● Enable Business Analysts to leverage ML ● Simplify the entire Machine Learning process Data scientists + Business analysts leveraging Automated ML
- 35. Confidential. ©2018 DataRobot, Inc. – All rights reserved Pervasive Machine Learning for greater customer experience and products Targeting Right product, Right time Tailored messaging Fair price Fast registration process Fair risk assessment Automated compliance / KYC Efficient support service Automated routing Chatbots Cross-sell relevant products Personalised offers recommendations Personalised Robo-advisor Trading strategies Action recommendations Security Compliance Forecasting Fraud Attrition and CV screening => 100’s of models to build and maintain => Requires automation!
- 36. The world’s most advanced Automated Machine Learning platform INSURANCE HEALTHCARE BANKING & FINTECH CYBER-SECURITY AND MORE 200+ 800,000,000+ Models built on DataRobot cloud #1 ranked Data scientists 4 50+Top 3 finishes Data scientists & Engineers (of 500+) $225M+ In funding 2012Founded HQ in Boston, MA
- 37. Confidential. ©2018 DataRobot, Inc. – All rights reserved