Munawar Khan

Deploys and maintains LLM-powered features end-to-end while designing RESTful APIs with FastAPI and Docker. Integrates OpenAI and Anthropic models into production pipelines and builds AI workflows using LangChain and LlamaIndex. Collaborates with domain experts to translate business requirements into scalable, deployable LLM features with measurable outcomes.

Conducted RLHF by evaluating and annotating AI agent execution traces across ML competition datasets, developing methodologies to identify factual errors, hallucinations, and reasoning violations in AI-generated code. Created detailed error taxonomies covering dataset hallucination, ungrounded reasoning result fabrication and logic inconsistencies to benchmark LLM performance on MLE-bench tasks.

Developed automated data pipelines and Python analytics solutions for healthcare and insurance clients. They translated clinical needs into technical tools, improved data quality, and ensured model fairness. Their work supported data-driven decisions and dashboards through close team collaboration.

Applied machine learning to solve financial challenges, developing models to predict credit risk and identify data patterns. This involved pre-processing data and feature engineering to create key risk indicators. The work translated complex findings into strategic recommendations and dashboards to support business decisions.