Cloud systems and AI infrastructure.

I design and defend architectures for machine learning systems in production — retrieval, vector search, data platforms, and the parts of a system that only become interesting once real traffic hits them.

• Retrieval-augmented generation: where it earns its cost, where a smaller index and a better question would have done.
• Vector search and the retrieval layer underneath it — recall, latency, and what degrades first.
• Regulated environments, including systems handling health data, where the compliance boundary shapes the architecture rather than decorating it.
• Platform and infrastructure design on Google Cloud.

I care about designs that remain useful after the novelty has worn off, and about being able to say which trade-off was taken and what it cost.