Computer Science undergraduate, NIT Andhra Pradesh
Varanasi Sathveek
I'm a Computer Science undergraduate at NIT Andhra Pradesh (B.Tech, 2023–2027). My work centers on machine learning and its internals.
I build retrieval-augmented systems over scientific literature and run mechanistic interpretability experiments on transformer language models, with a manuscript in preparation. Alongside research, I've shipped full-stack data platforms through internships at IIT Ropar, NIT Andhra Pradesh, and Intants.
Research interests
- Mechanistic interpretability of transformers
- Retrieval-augmented generation
- Natural language processing
- Graph-based analysis of relational data
- Applied machine learning
Selected work
All projects →MediLens — Biomedical Question Answering
A retrieval-augmented pipeline over PubMed: 262K abstract chunks embedded with a PubMedBERT encoder and indexed in FAISS, with fine-tuned BioBERT/PubMedBERT NER and chained cross-encoder reranking and generation (BioGPT/Flan-T5, local models via Ollama) returning structured JSON. Deployed as a Dockerized FastAPI service.
PyTorchHuggingFaceFAISSFastAPIDockerStudent Performance Analysis
End-to-end EDA on student demographic and academic data, engineering features and identifying key drivers of exam performance, then a modular ML pipeline comparing regression models — the best deployed as a Flask app for on-demand score prediction.
PythonPandasScikit-learnCatBoostFlaskSmart Environment Dashboard
A full-stack system for real-time monitoring of environmental sensor data, with interactive dashboards of parameter trends, threshold-based alerting, and device-management workflows that convert raw sensor streams into actionable signals.
TypeScriptReactNode.jsREST APIs