At Google, I designed a Go/gRPC routing layer for an internal Google Cloud operations service and introduced Redis caching that lowered p99 request latency from 42 ms to 16 ms in matched load tests.
I also developed a Python/FastAPI retrieval service for an incident-assistance prototype, preparing runbooks and resolved tickets for semantic search with LlamaIndex and Pinecone. With LangGraph and MCP, I orchestrated read-only diagnostic steps and evaluated OpenAI API and Anthropic Claude on anonymized incident scenarios.
At Yandex, I developed Python processing components for recommendation-data pipelines across 200+ nodes and approximately 10 PiB of data. Targeted C++ optimization reduced CPU time by about 50%, and tmpfs tuning reduced memory consumption by 20%.

