Shiv Sunder

I’m an engineer-turned-ML practitioner with a proven ability to learn fast and ship faster. I’ve worked across the full ML stack—from CV and NLP to LLMs, LangChain, and MLOps—while keeping quality and deployment readiness at the center.

In my Machine Learning Intern role at Quantiphi, I delivered 3+ end-to-end ML proof-of-concepts across CV, NLP, and Time Series. I optimized model pipelines with hyperparameter tuning, fine-tuned YOLO for real-time logo detection in live cricket broadcasts, and improved LLM response quality by building prompt tuning and retrieval validation pipelines to reduce hallucinations.

Through my projects, I’ve built a multimodal RAG system for PDF and scanned document ingestion (OCR + structured chunking) using hybrid FAISS + BM25 retrieval with reranking, plus a conversational QA system with persistent session context. I also developed a CNN-based License Plate Detection system using EasyOCR for reliable text extraction across varied conditions.