Milan Shrestha

I am an AI engineer focused on building GenAI-powered medical document processing systems that prioritize accuracy, scalability, and HIPAA-aligned workflows. I engineered OCR pipelines, custom indexing, and fine-tuned LLMs (3B and 8B parameters) to enable precise clinical summarization and deployed a GPU-optimized, asynchronous document-processing chatbot that raised throughput by 70%.

My work includes creating domain-specific datasets, implementing LoRA fine-tuning and model quantization to reduce computational overhead, and improving retrieval recall from ~60% to over 95% through structured query transformation and dynamic indexing. I combine practical engineering—using tools like RabbitMQ, LangChain, LlamaIndex, ChromaDB, and FastAPI—with strong collaboration and project management to deliver production-ready medical AI solutions.