Soumil Malik

I’m a CS undergraduate specializing in Agentic AI and LLMOps, focused on building autonomous multi-agent systems, distributed AI backends, and production RAG pipelines. I work across the GenAI stack—LangGraph, CrewAI, MCP, LangSmith, vector databases, and async task queues—to help models reason, act, and produce structured outputs beyond simple chat.

In my AI/ML Engineer Intern role, I engineered a teacher–student distillation pipeline to bring SOTA Qwen-Image-Edit reasoning into a Stable Diffusion 1.5 + LoRA workflow, and I built a synthetic data factory using a 4-bit quantized teacher model achieving 10× lower latency and 75% VRAM reduction (16GB → 4GB). In my projects, I’ve built an autonomous research platform with LangGraph + citations, a multi-agent GitHub code review system with tracing via LangSmith, and a multi-server MCP finance assistant with FastAPI + SSE—always emphasizing scalable execution, observability, and production-grade reliability.