At Pretectum, I build AI-powered data and full-stack systems, including a zero-downtime migration from AWS QLDB to Aurora PostgreSQL using AWS Glue and Kinesis CDC.
I've also built a BYOL AI gateway for multiple LLM providers, with encrypted key storage, model routing, token-usage monitoring, and cost dashboards. My AI pipelines for automated tagging, schema mapping, and Text2SQL reduced manual processing by approximately 40% and improved classification accuracy by over 25%.
Previously, at Somhako and Shink, I improved recruiter workflows, payments, matching models, and API performance. I enjoy building with TypeScript, Python, React, GraphQL, LangChain, and cloud data architectures, and I contribute to open source projects including Tracardi, SPCL-ETH Zurich, and Genomehubs.
