Code
Things I’ve built or contributed to.
Projects
Physics reasoning model A sample-efficient model trained to reason about introductory physics problems using symbolic constraints alongside learned representations.
Low-resource LM training pipeline A modular pipeline for training language models on small, morphologically complex corpora. Handles data collection, custom tokenizer training, and fine-tuning on Tulu and Sanskrit text.
AI agent for sales call analysis Built during an ML internship, an agent pipeline that transcribes, segments, and analyzes sales call recordings to surface insights for sales teams.
Healthcare RAG pipeline Built during a data science internship, a retrieval-augmented generation system for querying and surfacing information from clinical documents.