Now
What I’m focused on right now
A snapshot of current work, learning, and experiments — inspired by the /now movement. Updated as things change.
Last updated · July 2026
Building
- Production RAG + agent services on AWS ECS Fargate at EcoRatings.
- LLM evaluation harnesses that gate releases on correctness and hallucination risk.
- Fallback model routing across OpenAI, Anthropic, and open-weight models.
Learning
- Deeper graph-RAG and multi-hop retrieval strategies.
- Agent evaluation methodology beyond single-turn correctness.
- Cost/latency/accuracy trade-offs across model providers at scale.
Reading
- Papers on retrieval, reasoning traces, and LLM evaluation.
- Engineering blogs from Anthropic, OpenAI, and Stripe.
- Distributed systems fundamentals — the substrate under all of it.
Experimenting
- An AI playground for this site (retrieval viz, token streaming, reasoning traces).
- MCP-based tooling for agent workflows.
- Writing more — turning production lessons into public notes.