I am a computer science student at Karunya Institute of Technology and Sciences. I work on applied machine learning, LLM applications, and web apps.

My research interests are medical AI, explainable vision models, fairness-aware machine learning, and retrieval-augmented generation.

adni_project: Fairness- and horizon-audited sequential modality acquisition for Alzheimer's risk stratification, using real ADNI data (1,566 MCI-at-baseline subjects, October 2026 release). Published acquisition policies do no better than random acquisition at the same cost, while SPFA (signature MRI features plus pooled training) beats cost-matched random using MRI alone, at about a tenth of the cost of all four modalities. Python.

The four acquisition blocks it chooses between, with what each costs and adds:

fungi: Adaptive Confidence-Guided Multi-Scale Fusion (ACMF) for microscopic fungal species classification on the DeFungi dataset. It weights the predictions of 224px and 384px ViT-B/16 models by per-image prediction entropy. Written up as an IEEE conference submission. Python.

Radix-TMH: Hackathon project that tells a candidate how ready they are for a specific job. It covers JD analysis, resume parsing, profile building, and matching against a company's bar or a single JD. React, FastAPI, Supabase.

Drone-site: A drone rental site where customers browse the fleet, book a drone by the hour, and return it from their rental history, with the cost calculated on return. Next.js, Django REST Framework, PostgreSQL, Docker Compose.

ecommerce-gpt-chatbot: Retrieval-augmented chatbot running on a local LLM through Ollama. An LLM-based intent detector routes each query to FAQ search, product recommendation, product QA, or review sentiment. Jupyter, Sentence Transformers, Hugging Face.

gpt4_exp4: Console app that builds a learner profile, matches it against a catalog with Sentence Transformers, and has an LLM rank and explain recommendations, refining them from plain-English feedback. Python.

The rest are on my repositories tab.

Updated daily from the GitHub API.

• Languages: Python, JavaScript, TypeScript, HTML, CSS
• ML: PyTorch, Hugging Face, scikit-learn, pandas, Jupyter, Ollama
• Web and tooling: React, Next.js, Vite, FastAPI, Django, PostgreSQL, Supabase, Flutter, Docker, GitHub Actions, Git