🧠 CS Junior @ BITS Pilani Goa • Systems + AI Researcher • Infra-Aware ML Builder
I'm exploring the intersection of LLMs, HPC I/O optimization, and systems research. My current focus is on building trustworthy diagnostics for I/O logs using retrieval-augmented LLMs, and applying multimodal learning for intent discovery in livestreams. I enjoy debugging hard problems, both in code and compute.
-
🧩 Domain-Specialized RAG Systems
Built a pipeline to analyze Darshan logs using RAG + code generation + context retrieval (CSV+summary) → dynamic LLM diagnostics for I/O inefficiencies. -
🛠️ Custom Architectures for CIFAR
Training DiT-style networks without convs or transformers, achieving >95% accuracy using sliding window attention + class conditioning. -
🔭 Research Implementation: NAACL '22
Implementing Multimodal Intent Discovery using joint embeddings + clustering on livestream transcripts and visual frames. -
⚙️ Blockchain-Backed Platforms
Built a crowdsourced reporting tool with Django + React + Ethereum, integrating smart contracts and decentralized trust.
- LLMs for System Optimization
- In-Context Learning + Retrieval Techniques
- HPC Performance Tools (Darshan, Drishti)
- Low-level I/O Pattern Analysis
- MLOps, Infra-aware Model Design
- RAG + Code-gen Agents for Logs & Metrics
Languages: Python | C | Solidity | Bash
Frameworks: PyTorch | FastAPI | React | Django | Next.js
Infra: PyDarshan | Drishti | FAISS | Weights & Biases
Tooling: Git | Docker | VSCode | Vite | Tailwind | Linux