At Turbolab Technologies, I redesigned the LLM-based review analysis system as an event-driven pipeline that processes ~200k reviews per day, improving processing efficiency by 30% and cutting operational costs by 40%. I also built an evaluation framework to select prompts and models, and MCP servers that let AI agents operate infrastructure spanning 20+ services.
I maintain infrastructure for web scraping 1M+ records per day and have worked on PostgreSQL replication, service synchronization, and consolidating production systems. In my projects, I built Axidort, which scores websites for AI purchasing readiness, and AgentHire, a multi-agent marketplace using BeeAI and the A2A protocol.
