Cost Transparency and Model ROI
summary
Intuit runs 400+ models with 8,000 features on AWS using SageMaker, Kubernetes and Argo, and implements per-model asset IDs, fine‑grained tagging and BU chargebacks so each model’s storage, training and inference costs are traceable. Automated cost-testing runs ~40–50 hardware/config combinations to find optimal configs; smart defaults (e.g., 3-hour notebook auto-stop) and project-level resource isolation cut waste. Result: 15x ML usage growth without proportional cost growth and near-linear cost scaling with increased AI adoption.
AI-written summary · about
chapters (7)
- 0:00Introduction and background
- 4:59Cost transparency importance
- 10:21Technical cost control measures
- 18:34Non-technical transparency strategies
- 22:32Determining model ROI
- 24:59Outcomes and cultural impact
- 26:25Q&A session
related talks
26mManaging ML Models @ Scale
32mOptimizing AI/ML workloads for sustainability
28mAccelerating AI at Scale with Data
41mDeploy ML models for inference at high performance & low cost
19mResource Management Innovation for AI Workloads in Cloud Computing
27mScaling ML to the next level: Hosting thousands of models on SageMaker
29mMachine learning in production
20mSustainability architecture design for ML workloads