Notes on AI in production.
01
Agents are not features — they are systems
Going agentic isn’t about prompts. It’s about tools, state, constraints, logs, cost, and measurable quality.
02
Multimodal AI: from content to operations
Text, docs, vision, and voice create value only when connected to workflows and governance.
03
RAG you can audit
Great RAG isn’t “retrieve and answer.” It’s knowledge versioning, internal references, and measurable quality.
04
LLMOps: the production checklist
Without evals, regression checks, and observability, systems silently degrade. Production requires discipline.
05
AI FinOps: cost per tenant, not per vibe
Real control means cost per client, per feature, per flow—plus limits and fallback design.
06
Governance by design: audit trails for AI
Privacy and compliance aren’t paperwork. They’re architecture: logs, minimization, retention, and rationale.