Projects

Architected enterprise-grade retrieval-augmented generation (RAG) system processing 1,000+ technical service documents with hybrid search achieving 95% accuracy in error code resolution and component troubleshooting. Built scalable vector database architecture using FAISS with sentence-transformer embeddings.

HackIIIT Winner - Engineered enterprise-grade email AI assistant processing 10,000+ Gmail threads with 95% accuracy in event extraction and academic schedule parsing. Built production-ready OAuth2.0 authentication pipeline reducing manual email processing by 85%.

Megathon project - Designed novel BERT-inspired encoder achieving 87% accuracy in soft skill requirement extraction from job descriptions. Led ML model development processing 500+ job descriptions with 23% improvement over baseline matching.

ByteHacks 2nd place - Built production-grade blockchain from scratch with Node.js managing 100+ patient records. Developed predictive ML model achieving 89% accuracy in hurricane casualty prediction using custom-engineered Hurricane Katrina dataset.

Intel AI-Thon project - Created enterprise OCR system processing handwritten exams with 92% character recognition accuracy using CNN with CTC loss optimization. Implemented advanced NLP grading engine achieving 85% correlation with human teacher scores.

Research prototype - Designed scalable IoT architecture processing 8 concurrent sensor streams with 99.5% uptime using Arduino Nano BLE. Built responsive Flutter application displaying real-time sensor data with sub-100ms latency.