AI Insights
Practical perspectives on agentic AI.
On-device inference, agent architecture, and shipping LLM features to production — from the engineering side of the table.
What we write about
Four areas we keep coming back to.
Agentic system design
Intent routing, tool use, memory management, multi-model orchestration. The architectural decisions that determine whether an agent is useful or chaotic.
On-device and edge AI
llama.cpp internals, quantization tradeoffs, multimodal projectors, streaming token output. When and why to run models off the cloud.
Production AI engineering
Observability, evaluation harnesses, fallback strategies, cost discipline, and the operational rituals that keep AI features alive past launch week.
Lessons from the field
Case studies and post-mortems from real engagements — the specific bugs, the specific fixes, the things we wish we’d known sooner.
Latest posts
Recent writing.
Stay in the loop
Want these in your inbox?
Roughly one piece per week. No marketing fluff — just notes from the work.