Creating an IDP for AI Engineering - Max Körbächer
summary
Argues for an AI internal development platform that extends an existing IDP rather than copying models claims the competitive edge is the platform not the model and cites the MAT landscape with 1,150 tools to show scale details actual tools to integrate: MLflow, Kubeflow, vLLM, LlamaIndex, LangChain and LightLLM describes concrete features to build: semantic routing, model versioning, cost controls for GPUs, observability and guardrails points to operational examples (Uber Michelangelo, Netflix Metaflow) and prescribes a phased rollout: automate, enable self-service, then add governance via an AI developer portal.
AI-written summary · about
chapters (15)
- 0:00Introducing speaker and topic
- 1:46AI adoption challenges
- 3:53AI tool landscape explosion
- 5:38LLM era complexity
- 7:43Integration tax and obsolescence
- 8:47Three platform engineering principles
- 11:17Governance and observability
- 13:46Model management challenges
- 16:16AI developer portal
- 19:03Real-world AI platforms
- 20:27Benefits of AIDP
- 22:36Thinnest viable IDP
- 23:27Q&A: AIDP vs IDP
- 25:33Q&A: Management benefits
- 28:42Q&A: Model fine-tuning
related talks
24mSoftware 3.0 and the AI Engineering Landscape
28mTransforming Internal Developer Platforms with AI: Accelerating Innovation
31mAgentic AI
47mEmbedding AI: Don’t Burn Your Fingers in the Heat of the AI revolution
29mLeveraging AI to Enhance DevEx & Operational Efficiency
19mDeveloper Centric AI
26mArchitecting Agentic Development Platforms
30mRab Ne Bana Di Jodi - Why Platforms & AI Need Each Other