Prototypes to products: Build & deploy at-scale DS driven products
summary
A practical framework for taking ML prototypes to production at scale, illustrated by Envestnet | Yodlee’s Transaction Data Enrichment (TDE) product that handles millions of financial transactions daily. Covers concrete stages and technologies: data gathering/cleaning/normalization, integrating domain expertise, model selection and training (including modern methods like deep learning), evaluation, architecture and design choices for at-scale deployment and redundancy, continuous monitoring for model drift, and logging customer feedback.
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:21Introduction Slide
- 15:23Steps when data is involved in decision making
- 16:51Keys steps for a deployment ready Al/ML solution
- 25:24Systematic Data Science Strategy for Scalable Foundation
- 32:08Domain Expertise + Model Flexibility
- 34:31Guess the number of dusters
- 35:29Human leaming is more flexible
- 43:05UI/UX to camouflage ML shortcomings
related talks
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
47mDeveloping a digital product. What it is and how
42mMasterclass: The right path to ML
32mTeam Teardown: Evolving an ML Platform
28mTaking Data Science in industry from zero to production
19mDevDataOps - Data-Driven Product Development for Software Engineers
26mGoing all-in on data products
50mMasterclass: Building a data stack in 2021