I work on the infrastructure layer between models and products.

Calling a model is usually not the hard part. The hard part is everything around it. Agent loops, retrieval, streaming media, traces, latency budgets, control planes, and the distributed services that still have to work when real users show up.

That’s where my work has been going real-time AI, voice infrastructure, agents, retrieval, observability, and the systems needed to make machine learning useful in real products.

Models will keep getting better. The harder question is how to make them reliable, fast, debuggable, and useful in real systems.

If any of this resonates, feel free to read through my writing, take a look at what I'm currently working on, or just reach out over email if there's anything I can help with.