Project (Spam email detection) at Spam Email Detection

Built an end-to-end spam classifier using TF-IDF vectorization and Naive Bayes. Implemented a complete NLP pipeline including tokenization, stop-word removal, and stemming. Packaged the model as a reusable module with a clean predict() API for drop-in integration with an email backend.

Project (Backend + Frontend; Carbon credits, digital twin, crisis detection) at Bharat-Setu

Architected a zero-cost, pilot-ready platform with three modules: Carbon-Kosh (satellite-verified carbon credits), Gram-Twin (village digital twin), and Migration-Shield (boolean crisis detection). Replaced 8 AWS services with open-source equivalents, modeled data using 6 SQLAlchemy models, implemented 10 business logic services, and created grouped REST API routes. Added an automated fraud protocol with traffic-light thresholds and account controls, shipped 76 unit tests across 4 suites, and built a mobile-first farmer PWA with a push-to-talk voice interface plus a React admin dashboard.

Open-Source Contributor (Privacy-first AI workspace) at PrivGPT Studio

Built a privacy-first AI workspace supporting local and cloud LLMs with offline fallback. Implemented multimodal cross-referencing to improve contextual accuracy. Enabled uploads of PDFs, videos, images, and audio with transcription, integrating LLM workflows for better context retrieval and user interaction.