Rishit Pant

I'm pursuing a B.S. in Data Science & Applications at IIT Madras, and I focus on turning AI/ML ideas into working, production-grade systems rather than notebooks.

As a Research Intern at IIT Hyderabad, I built a PyTorch deep learning model using Federated Learning for medical image classification, benchmarking custom aggregation strategies against FedAvg and reducing aggregation workload by 38% while preserving accuracy.

More recently, I built AskTheHandbook, a production RAG chatbot that grounds LLM answers in official documents using hybrid retrieval (BM25 + ChromaDB) and cross-encoder reranking. I spent real time on the parts that don't show up in tutorials — cleaning messy source data, addingretry/backoff logic for API failures, and building an evaluation framework (DeepEval, Langfuse) to catch hallucinations and track latency and cost before deploying anything.

I also built System Threat Forecaster, an end-to-end ML pipeline that predicts malware infection from endpoint data, where I benchmarked multiple algorithms, reduced the feature space through statistical analysis, and applied MLOps practices (MLflow, CI/CD) to keep it reproducible.

I'm most interested in roles where I can build things that hold up in production, not just in a demo.