Reading risk out of noise — in climate data and in candidate pools.
M.Sc. AI & ML student at IIIT Lucknow, building climate models at the Climate Resilience Observatory (UP Govt.) and shipping full-stack AI platforms — from satellite-driven risk scoring across 75 districts to CPU-only ranking of 100,000+ candidate profiles.
I'm an M.Sc. AI & ML student at IIIT Lucknow (CGPA 8.8), currently working as a Data Science Intern at the Climate Resilience Observatory, UP Government, where I process over a decade of ERA5 climate data to predict district-level rainfall and contribute to watershed management systems.
Outside of that, I build full-stack AI platforms end to end — from a FAISS-powered candidate ranking engine that processes six-figure applicant pools on CPU alone, to a satellite-driven urban vulnerability index built on real NASA and ESA data.
I care about systems that turn raw, noisy data — satellite rasters, applicant pools, medical scans — into something a person can actually act on. That's the thread across everything I build: prediction is only useful when it's legible.
Currently deepening my focus on NLP, reinforcement learning, and MLOps as part of my coursework, while looking for opportunities to ship GenAI products with real users.
Grouped by how I actually use them — not an exhaustive keyword dump.
Three systems, three kinds of noisy data — applicant pools, satellite rasters, medical scans.
Pulled live from the GitHub API on page load.