At Google DeepMind, I led a Gemini applied-AI workstream that delivered production LLM capabilities for Google Cloud enterprise customers. I directed the launch of Gemini-powered enterprise search, which generated $28M in incremental ARR during its first year.
I cut serving costs by 46% through speculative decoding, KV-cache quantization, and workload-aware batching on TPU v5e. I also introduced an evaluation framework that reached 91% agreement with specialist reviewers.
Previously, at Meta AI, I improved Facebook News Feed candidate generation with a two-tower PyTorch retrieval model that increased candidate recall by 12.4%. At Payhawk and Zalando, I worked on full-stack financial and commerce platforms, and I began my career building customer-facing web interfaces at SoftServe Global.

