Software Engineer at Atreyus LLC

Engineered an agentic vector store pipeline that delivered 93% chunk-level accuracy for retrieval. Applied prompt engineering to boost workflow efficiency and improve model estimation accuracy by 10%, ensuring closer alignment with expected results. Used ARQ and AWS ECR to orchestrate large file processing, increasing throughput and reducing processing time by up to 20x. Optimized embedding pipeline, achieving a 9% increase in performance and 3x processing speed. Developed an evaluation pipeline for a RAG system, resulting in 91% chunk accuracy.