At Tredence, I build backend and MLOps systems for a Fortune 500 CPG client, including a FastAPI service powering a multi-agent LangGraph text-to-SQL platform used by 150+ analysts, stakeholders, and field staff.
I reduced end-to-end query latency by 36% and LLM token costs by 25%+ through semantic context filtering and response caching. I also built reusable Databricks tooling that reduced MLflow experiment setup from 60+ lines to four.
Previously at Asgard.World, I built an asynchronous Python pipeline using S3, SNS, SQS, and Lambda to process 500K+ product listings daily, alongside a CLIP-based product classification and semantic search system.
I also built a Go YANG-to-Protocol Buffers compiler at Altiostar and created loggergo, a structured logging pipeline using Redis, PostgreSQL, and Grafana. I enjoy solving production bottlenecks through focused benchmarking and reliable infrastructure design.
