CS undergrad (B.Tech, 3rd year) specializing in AI & Machine Learning β I build production-ready intelligent systems from scratch.
- π Β Shipped two deep learning systems at 98β99% accuracy and a full multimodal RAG pipeline
- π§ Β Currently deep in LLMs, Agentic AI & RAG architectures
- π οΈ Β Passionate about end-to-end ML β from dataset curation to deployment
- π₯ Β Open to collaborating on AI/ML open source projects
Languages
AI / ML & GenAI
Vector DBs & Embeddings
Databases & DevOps
Eliminated manual product cataloguing with an image-to-text pipeline (vision model β FAISS embeddings) that auto-generates searchable product descriptions. Replaced keyword search with natural-language product discovery across a multimodal vector store. Delivered a fully deployed Streamlit app with zero friction for non-technical users.
RAG FAISS Multimodal LangChain Streamlit
Achieved 99% weapon-detection accuracy (guns, knives) on custom CCTV datasets using YOLOv8 with targeted data augmentation. Fine-tuned an RS3D spatiotemporal model for 83% violence-classification accuracy, outperforming baseline. Built a real-time dual-model inference pipeline with OpenCV + ffmpeg for live frame-level threat flagging.
YOLOv8 RS3D OpenCV Real-time Inference Streamlit




