At RoundtableCI, I designed and built a dynamic model-routing platform that selects a single model or multi-agent pipeline for each query, cutting inference cost and latency by ~43%. I own the distributed backend and orchestration layer, including architecture decisions, database design, and LangChain-based routing logic.
At CloveCI, I’m the technical lead across a five-engineer team building four products. PrepHelp provides chat-with-PDF and syllabus-aligned Q&A and has 300+ active users; I also set technical direction and review engineering decisions across the products.
On ContextIQ, I built an explainable RAG system with a four-agent pipeline that streams its reasoning trace via SSE. On Athena, I benchmarked and tuned a dynamic model router that cut inference costs 80% in testing while holding a fixed quality threshold.
I’m the primary author of “Beyond Retrieval: Execution Intelligence for Hierarchical Agentic Retrieval,” submitted to ICCMM-26. Validated through ContextIQ, the work benchmarked ~2.6x faster than a production RAG baseline, with a higher success rate.
