Priyank Agrawal is an AI Engineer based in Gwalior, India, who builds production LLM agents on AWS Bedrock and LangGraph, with backend systems engineering underneath them. He has 3+ years of professional experience and works at Planet Sustech, where he built a single-agent ESG system with 21 typed tools. He is also the author of neverempty, an open-source Python eval harness on PyPI that measures how often a tool-calling agent tells a user "no results" when a tool actually failed. He is on GitHub as Priyank032.

neverempty is an open-source Python library by Priyank Agrawal, published on PyPI (pip install neverempty). It makes tools return Ok, Empty or Err instead of a bare list, tells the model explicitly when a tool failed, and injects tool faults on purpose to measure how often an agent misreports a failure as "no results". It reports Wilson confidence intervals and gates CI with an exact McNemar test. Its headline metric has not yet been measured on a production agent.

Priyank specializes in LangGraph, LangChain, AWS Bedrock, Python, Node.js, FastAPI, Next.js, MongoDB, PostgreSQL, Redis, AWS Lambda, and Docker. He builds RAG pipelines on vector search (MongoDB Atlas, pgvector), typed tool-calling agents, and agent evaluation.

Yes, Priyank is open to new opportunities and collaborations. He is available for remote or hybrid roles and can join within 2-4 weeks. He is also open to freelance and contract work. Contact him at priyankagrawal76660@gmail.com.

Priyank built an ESG agent on AWS Bedrock and LangGraph — deliberately single-agent with 21 typed tools, role-scoped tool binding, and server-derived figures — which cut manual ESG analysis time by roughly 70%. He is the largest contributor to that ~240K LOC platform, where he took a dashboard query from over three seconds to under one by removing 14 $lookup and 8 $unwind stages and cutting documents scanned from ~38K to ~3.2K. He also built YojanaKhoj, matching citizens to 101 government schemes via a three-valued rule engine.