Jash Shah
I am a B.Tech graduate from Dhirubhai Ambani University in Information and Communication Technology (Oct 2022 – May 2026).
I am joining Factset as a Machine Learning Engineer I (AI Enablement) in Hyderabad, where I will be working on entity-graph enrichment pipelines for news articles, SEC filings, and broker research.
Most recently, I was an AI Engineer Intern at Binocs Labs, where I designed production RAG pipelines with HyDE and DecomposeQuery transformations, LLM reranking, and citations, and self-hosted Arize Phoenix for LLM observability across agentic workflows. I also contributed to Arize Phoenix (9.5k+ stars) and OpenInference as an open source contributor.
I am motivated by a central question: how do we know when AI models are genuinely improving? My research in AI evaluation aims to build more trustworthy and meaningful benchmarks for large language models. I conduct research at the Knowledge and Discovery Lab under Prof. Sourish Dasgupta, in collaboration with Oak Ridge National Laboratory, on DISCERN — a diagnostic framework for epistemic non-triviality in LLM-based hypothesis generation (under review at ACL 2026).
I served as an Undergraduate Teaching Assistant for the IE406 Machine Learning course under Prof. Pritam Anand and was the Convenor of the AI Club at DAU. I mentored 100+ students through a course called AI Odyssey, and co-organized Devolution as a core team member of Google Developer Group, On Campus DAIICT.
I work with Python, C, C++, SQL, FastAPI, Django, Flask, PyTorch, TensorFlow, LangChain, LangGraph, LlamaIndex, Google ADK, n8n, AWS, Docker, Kubernetes, and vector databases including Pinecone, Chroma, and FAISS.