LN Renganarayana, Architect, ML Platform and Services, Workday San Francisco
summary
Workday builds enterprise ML products that initially overpromises—an early push for 40 apps yields one usable MVP—because k‑fold shows 95% accuracy but temporal validation drops it, revealing validation choice matters. A retention‑risk product fails adoption (<10 customers) until results expose actionable factors and training is provided. Enterprise constraints forbid pooling customer data, forcing tens of thousands of per‑customer models and automated pipelines for build/validate/version/deploy. Platform includes APIs, ML services, persistence and a DSL called expresso to let domain experts prototype and productionize models.
AI-written summary · about
chapters (6)
- 0:00Start
- 1:04Professional Life at Workday
- 1:38Pace of Innovation
- 5:31Do Spend the Time in Educating Your Stakeholders
- 12:42How Do You Scale Machine Learning for Enterprise
- 13:00What Are the Dimensions of Scale
related talks
31mHow Workday Fine-Tunes LLMs at Scale
23mEd Fullman On Stage
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
31mUse-Case Driven ML Platforms
14mScaling ML platforms to industrialize AI/ML development
23mML Platform Challenges & Applications at Freshworks
27mML Platform Journey
05mInterview