hi. Prateek here.

I'm an AI systems engineer focused on making large language models work reliably in production. That means distributed training infrastructure, inference optimisation, and the retrieval architectures that sit underneath LLM applications. I care about the systems layer — the part between research papers and deployed services that rarely gets written about.

My path through this field has been deliberately cross-cutting. I've worked on multi-node GPU training clusters (FSDP, NCCL, Slurm), built high-throughput backend APIs that serve ML predictions under load, and published research on applied deep learning. The thread connecting all of it is a preference for understanding systems end-to-end rather than specialising in a single layer.

Currently I'm building two open-source tools: Setu, a pre-emptive discipline layer for AI coding agents, and RAG Arena, a competitive evaluation framework for retrieval-augmented generation pipelines. I'm also open to fractional AI systems leadership and consulting engagements for teams scaling their ML infrastructure.

I completed my MS in Computer Science at the University of Florida and my BTech in CSE at BML Munjal University. NVIDIA certified in Deep Learning, CUDA, and Generative AI with Diffusion Models.