Gap-filling missing TROPOMI satellite data across India (2019–2024) for NO₂, SO₂, CO, PM2.5, and PM10 using ML-based imputation to enable continuous air-quality monitoring.

• →Built ML pipeline to fill TROPOMI gaps via DINEOF approach — internal cross-validation r = 0.8
• →Validated aggregated TROPOMI outputs against OMI at matching resolution (r = 0.53)
• →Identified Indo-Pacific pollution hotspots from multi-year reconstructed time series
• →Ongoing: downscaling TROPOMI retrievals to fine spatial resolution using deep learning

Built an offline AI document QA system for defense use — enabling secure, air-gapped querying of classified PDF and DOCX documents without any external API calls.

• →Designed RAG pipeline: PyMuPDF ingestion → ChromaDB vector store → CrossEncoder reranking → LLaMA 3.2 via Ollama
• →Achieved ~40% retrieval accuracy improvement through chunking strategy and reranking
• →Reduced document load time by 50% via persistent embedding cache
• →Delivered Streamlit UI for non-technical defense staff — fully offline, zero data leakage

Selected as team lead after winning the internal college hackathon (1st out of 100 teams) to redesign and deploy the official MVSR college website.

• →Led redesign using HTML5, Tailwind CSS, and vanilla JS — dark/light mode, sticky announcements
• →Coordinated a cross-functional team for live rollout to the college community