At JunctionNet AI, I designed and delivered Python services that let non-technical decision makers generate agent-based models to address business questions. In pilot programs, this work cut time to the first model draft by 45%.
I built scalable FastAPI services and RAG pipelines using Pinecone and FAISS, and developed LLM evaluation and prompt-optimization workflows. The retrieval pipelines improved grounding and relevance across enterprise scenarios.
I also worked on model validation architecture and MLOps deployment, monitoring, and rollback across AWS and Azure. My work included performance improvements that reduced p95 response latency by 38% and automated testing practices that reduced production defects by 32%.
Before JunctionNet AI, I developed Python services and RAG prototypes at Nord Quantique, and built enterprise applications at Endava and HCL Enterprise. I enjoy turning modeling and simulation capabilities into software that decision makers can use.
