ROHIT UPADHYA

rohit@portfolio

—————————————————————

role: Software Engineer — ML @ Whatfix
focus: ML · LLMs · retrieval · agents
school: MSc CS, TU Munich
: BTech CS, NIT Silchar
based: Bangalore, India

about me view blog see research

featured

projects research

Master's Thesis

Cross-Lingual Query based Paragraph Retrieval for European Court of Human Rights Judgements

LexCLiPR: a cross-lingual dataset from ECtHR judgements, used to benchmark and fine-tune LMs (Siamese and Two-Tower setups), plus GAVEL, a graph-based approach to paragraph retrieval. Accepted to ACL'25.

GAVEL — Graph-Augmented Paragraph Retrieval for Legal Judgments

Extends LexCLiPR with dual Graph Attention Networks (local 5-hop adjacency + global BERTopic topic nodes) trained under contrastive loss, outperforming a fully fine-tuned bi-encoder with a frozen mDPR encoder.

SmoLLM — 109M Parameter Language Model from Scratch

A Llama-style decoder-only transformer built from scratch — custom BPE tokenizer, RoPE, RMSNorm, SwiGLU — pretrained on FineWeb-Edu, then instruction-tuned on databricks-dolly-15k.