Sahil Malla

I’m a Generative AI & ML Engineer with 3 years at Deloitte USI, focused on delivering production-grade LLM systems that go beyond prototypes. I build LLM-serving APIs and end-to-end RAG pipelines, with an emphasis on reliability, speed, and measurable impact.

In my role, I reduced model inference latency by ~35% across multiple production microservices by designing and deploying FastAPI-based LLM-serving endpoints. I also improved RAG retrieval accuracy by ~28% versus a keyword-search baseline using LangChain, ChromaDB, and sentence transformers, and increased pipeline throughput by ~40% through an event-driven architecture with RabbitMQ.

I’m hands-on with LLM fine-tuning (LoRA/QLoRA), agentic AI approaches, vector databases, and OpenAI/HuggingFace APIs. From building an Agentic RAG system with persistent context memory to grounding answers and reducing hallucinations with structured retrieval and re-ranking, I consistently aim to close the gap between prototype and production while collaborating effectively across teams.