Udit Jain

Udit Jain

Agentic AI Engineer @ Samsung Research AI Core Team

I am an AI engineer based in South Korea, with a foundation in Computer Science from IIT Delhi (B.Tech, CGPA 9.10/10.0). I work at the frontier of large language model post-training and autonomous agent systems.

At Samsung Research’s AI Core Team, I build and train cutting-edge agentic AI models. My work spans the full post-training lifecycle: designing RLHF and DPO pipelines, implementing SFT for instruction following, applying knowledge distillation to compress frontier-scale capabilities into deployable models, and building the evaluation infrastructure to measure agent performance on multi-turn benchmarks.

Before this role, I spent two years as a Computer Vision Engineer in Samsung’s Next Generation Digital Appliances division, where I shipped on-device food recognition models to millions of refrigerator users – achieving <40MB model size and <70ms inference on NPU hardware. Prior to that, I built Samsung Health’s research data platform and led large-scale analysis across 200M+ users’ health data.

My engineering philosophy: understanding the system end-to-end matters more than mastering any single tool. Whether it is training a reward model, debugging a distributed NeMo run, or profiling an NPU kernel, I want to understand why it works – not just that it does.

I am always open to discussing research ideas, technical collaborations, or new opportunities. Feel free to reach out.