Automated Machine Learning

Enabling the AI-Driven Enterprise

Transcript

1. Tema: Enabling the AI-Driven Enterprise with Automated Machine Learning
2. Hola! Soy Scott Bergquist Senior Director of Business Development DataRobot scott.bergquist@datarobot.com You can find me on Twitter @scbergquist
3. “Those who rule data will rule the entire world.” M A S A Y O S H I S O N
4. “The leaders in artificial intelligence will rule the world.” V L A D I M I R P U T I N
5. “Companies have to race to build AI or they will be made uncompetitive. Essentially, if your competitor is racing to build AI, they will crush you.” E L O N M U S K
6. Closing the AI Gap with Automated Machine Learning The global demand for machine learning solutions greatly exceeds the production capacity of all the data scientists in the world. Demand for machine learning & AI Data scientists in the world 2010 2014 2016 2018 2020 2022 2024 2008 2012
7. No longer a viable option Continuing to have only humans manually build Machine Learning models is
8. 10-100x Increased productivity for data scientists Democratization of data science: 1000x more AI Producers. We have been focused only on inventing and improving enterprise- grade automated machine learning since 2012. Demand for machine learning & AI Data scientists in the world 2010 2012 2014 2016 2018 2020 2022 2024 2008 The only viable solution: Automated Machine Learning
9. The world’s only truly automated machine learning platform DataRobot is better and faster than 99.9% of the world’s data scientists The company was founded with a specific mission: To teach machines to do data science Machine learning is about automation... DataRobot is about the automation of automation.
10. 200+ Data Scientists & Engineers (of 400+) 4 #1 ranked Data Scientists 50+ Top 3 finishes BANKING HEALTHCARE MANUFACTURING INSURANCE MANY MORE The world’s most advanced Enterprise Machine Learning platform 2012 Founded, HQ in Boston, MA $124M In funding 650,000,000+ Models built on DataRobot Cloud
11. Largest US Supermarket chain Largest US for-profit Healthcare System 7 of the Top 10 Global Banks World’s largest Retailer 3 of Top 5 global Reinsurers 2 of the worlds largest Biotechs 2 of Top 5 Global Telecom providers 3 Major League Baseball teams Federal & Public Sector Agencies Largest mobile payments app 2 of the largest Hedge Funds by AUM Largest US Pharmacy chain USED BY SOME OF THE WORLD’S MOST PROMINENT COMPANIES
12. $150 Million in annual revenue and cost savings achieved in under 12 months
13. Wealth Management (private client) • New customer prospecting • Relationship deepening • Customer attrition • Robo-advising Special Assets/Workout • Optimizing resolution strategy; scenario testing • Collection and recovery management • Valuation for securitization / secondary market Risk • BSA/AML • Fraud detection/prevention • Streamlining model risk management • Credit risk & Loss forecasting • Targeted Risk Review Wholesale/Commercial Banking • Sales and prospecting • Credit Risk/Scoring • Loan pricing optimization • Automated loan approval • Loan operations optimization • Collateral valuation • Early warning system Mortgage Company • Predict Prepayment Speed • Manage MSR value/hedge • Early warning of delinquencies • Prioritize collections Retail Banking • Targeted marketing • Churn Risk • Call center optimization • Complaint management • Credit risk modeling Investment Banking/Capital Markets • Margin management • Marketing, AI-based matchmaking • Offering/deal/event prediction and prospecting Cards • Application and transactional fraud detection • Offer targeting • Rewards design/marketing • Collections optimization • Call center optimization Merchant Services/POS • Card and merchant fraud detection • Marketing/lead optimization • Upsell/cross-sell, relationship deepening Sell-side/Markets • Development of trade recommendations, research • Trade surveillance, communication surveillance • Optimized trade settlement • Customer behavior/value added services Inst’l investment management • AML & surveillance • Model risk management • Attribution/behavior analysis • Trade settlement optimization • Collateral management • Sales and marketing THERE ARE HUNDREDS OF OPPORTUNITIES TO OPTIMIZE EVERY LINE OF BUSINESS IN A BANK (…or any business) Treasury and Cash Management • Smart accounts receivable • NSF early warning system/alerts • Cash flow forecasting • Transactional FX optimization
14. The AI Bottleneck: List of skills required to be a Data Scientist 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
16. List of skills needed for executing on AI opportunities by leveraging Automated Machine Learning: 1. Knowledge of the business and business problem 2. Knowledge of the data 3. Pragmatic education and mentoring
17. Automated creation of end-to-end workflows A blueprint is a combination of data cleansing, feature engineering, feature selection, pre-processing, hyperparameter tuning, machine learning algorithms, and more. Data Categorical Variables Numeric Variables Text Variables One-hot Encoding Univariate Credibility Estimates with Elastic Net Category Count Missing Value Imputation Converter for Text Mining AutoTuned Worn N-Gram Text modeler using token occurrences Search for Ratios Search for Differences Gradient Boosted Greedy Trees Classifier with Early Stopping Prediction Our Approach: The Blueprint
18. Key Differentiators Accessibility Speed Transparency Mass Model Production Accuracy Model Deployment Little to no data science experience required to get started. Train and test hundreds of models in a fraction of the time it takes the average data scientist to create one model. Machine learning is not a black box. DataRobot enables users to see everything that is happening under the hood. The speed of model creation turns your data scientist into a mass producing model factory! DataRobot has consistently delivered top results in international data science competitions across a variety of data sets. No re-coding required. Deploying models to production is a matter of minutes.
19. WHAT TO WATCH FOR IN THE DEMO Best practices and guardrails automatically applied Automation with flexibility for experts Full transparency. No black box models Automatic benchmarking/challenger models Automated Model Documentation Flexible deployments and automatic monitoring
20. GRACIAS! ANY QUESTIONS? Puedes encontrarme como: @scbergquist scott.bergquist@datarobot.com