Built a vision-aligned HR hiring platform. Moved beyond standard keyword filtering by engineering an LLM pipeline that dynamically asks indirect behavioral questions, parses complex CVs, and ranks candidates based on strict job/vision alignment.
HR teams were drowning in thousands of CVs, relying on basic keyword-matching ATS systems that failed to capture nuance. Hiring managers complained that candidates looked good on paper but completely lacked the behavioral alignment and specific technical vision required for the roles. Screenings were taking weeks, costing thousands in lost productivity.
We engineered a multi-stage LLM evaluation pipeline. Instead of dumb keyword extraction, the system:
Time-to-hire was reduced by 60%. The system achieved a 92% agreement rate with human HR directors on top-tier candidate selection, completely eliminating the manual screening phase and ensuring only highly-aligned visions made it to the interview stage.
gpt-4-turbo (Evaluation), text-embedding-3-small (Vectorization)Calculate Your MVP Budget
An automated academic intelligence system that scrapes, parses, and taxonomizes thousands of PDF research papers. Designed for a specialized consulting firm to accelerate their market research phase by 400% through high-accuracy automated tagging.
Architected a distributed reservation engine capable of handling 50,000+ simultaneous bookings for a national travel platform without a single millisecond of downtime.