Building the Next Generation of Agentic AI

Welcome! I am Udit Jain, an Agentic AI Engineer at the AI Core Team, Samsung Research, focused on the frontier of large language model post-training and autonomous agent systems.

What I Work On

My current work lives at the intersection of post-training research and production-scale AI:

• LLM Post-Training: RLHF, DPO, SFT pipelines for instruction-following and alignment
• Reinforcement Learning for LLMs: Reward modeling, policy optimization, and RL-based fine-tuning
• Knowledge Distillation: Transferring capabilities from frontier models to efficient, deployable targets
• Agentic Evaluation: Building multi-turn agent benchmarks and evaluation pipelines for complex reasoning tasks
• Training Infrastructure: Distributed training with NeMo, large-scale inference with vLLM

Previously, I led computer vision R&D for Samsung’s next-generation smart appliances, shipped on-device AI models to millions of users, and built health data pipelines over Samsung Health’s 200M+ user dataset.

Please explore my Projects, read about my Professional Journey, or view my Resume & Skills.