Topic hub
AI Platform Architecture
Practical notes on designing production AI platforms, LLM routing, self-hosted assistants, RAG systems, and the infrastructure patterns behind reliable AI products.
What this covers
AI platforms need more than model access. They need routing, observability, safety boundaries, cost controls, data access, and a clear operating model for how humans and machines collaborate.
Search intents answered
- How should a production AI platform separate model access, retrieval, orchestration, and user experience?
- What trade-offs matter when self-hosting an AI assistant inside an organisation?
- How do teams keep AI workloads observable, governed, and cost-aware?
- When does an AI assistant need a platform architecture instead of a single app integration?
Related insights
AI
Lessons Learnt Self-hosting an AI Assistant
A practical guide to self-hosting an AI assistant on Azure using OpenWebUI, Kubernetes, LiteLLM, and a custom RAG pipeline.
AI
The AI-Native Data Platform We've All Been Waiting For
Why an AI-native data stack matters and how Nvidia is building toward it.
AI
Stop Writing README.md. Start Writing AGENTS.md
Why agent-first documentation is becoming the practical way to guide AI Agents in real codebases.
Tech
The Infinite Canvas and the Finite Mind
A personal reflection on what happens when the tools outpace the thinker.
Related case studies and pages
Designing an AI platform?
Bring a real architecture decision to a mentorship session and work through the platform trade-offs.
Explore mentorship