AI systems that hold up
Designing model-backed workflows with observability, routing, latency budgets, and fallback paths.
- Gateway-level usage and cost tracking
- Prompt and response evaluation loops
- Operational UX for non-demo behavior
Learnings
What I am sharpening now.
A working log of the themes that keep showing up in my projects. This page replaces random notes with the areas I am actively improving.
Training and coaching products
Using sensors, context, and feedback loops to make fitness guidance feel timely and personal.
- Heart-rate zone coaching
- Strength training progression
- Conversational onboarding
Software design notes
Keeping systems small, legible, and easy to change without losing correctness at the edges.
- Fast feedback loops
- Readable architecture boundaries
- Root-cause fixes over patches
Writing in public
Turning project work into durable notes: what changed, what failed, and what I would do again.
- Build logs
- Decision records
- Post-launch reflections