AI engineer at Samsung Research's AI Core Team in Seoul, building agentic AI systems through LLM post-training, reinforcement learning, and knowledge distillation. IIT Delhi CS, CGPA 9.10.
Previously shipped on-device computer vision models to millions of Samsung smart appliance users and ran health data analysis across Samsung Health's 200M+ user dataset.
LLM Post-Training & Alignment — End-to-end SFT, RLHF, and DPO pipelines for instruction following and safety alignment of large language models.
Reinforcement Learning for LLMs — Reward modeling, policy optimization (PPO variants), process-level and outcome-level reward signals for improving reasoning, planning, and agentic task completion.
Knowledge Distillation — Transferring frontier-model capabilities to deployable student models while maintaining benchmark performance at lower compute and latency.
Agent Evaluation — Multi-turn evaluation pipelines and benchmark suites for tool use, long-context reasoning, and autonomous task completion at scale.
Training Infrastructure — Distributed training with NeMo on multi-node GPU clusters. Large-scale inference and evaluation with vLLM.
| Area | Technologies |
|---|---|
| LLM Training | NeMo, vLLM, PyTorch, Hugging Face Transformers, PEFT/LoRA |
| Methods | SFT, RLHF, DPO, PPO, reward modeling, knowledge distillation |
| Computer Vision | OpenCV, TensorFlow, object detection, segmentation, NPU optimization |
| Data & Cloud | AWS (S3, EMR, Glue, Athena), Spark, BigQuery, Apache Superset |
| Languages | Python, C/C++, Kotlin, Java |
| Systems | Distributed GPU clusters, Docker, Linux kernel, FPGA |
Agentic AI Engineer, AI Core Team (2026 - Present) Post-training pipelines, RL for LLMs, distillation, agent evaluation frameworks.
Computer Vision Engineer, Digital Appliances R&D (2023 - 2025) On-device food recognition for smart refrigerators — <40MB model, <70ms inference, >90% accuracy on NPU. Showcased at CES 2024. Product Video
Software Engineer, Data Service Lab (2022 - 2023) Built Samsung Health Research Application connecting devices to clinical research. Won UX Design Award 2023. Press Release
Data Science Engineer, Data Intelligence Lab (2020 - 2022) Health data analysis on 200M+ users (300TB on AWS). Global sleep study featured on Korean national TV and Samsung Newsroom.
- Harvard University, Visual Computing Group (2020) — Differentiable rendering, GPU-accelerated Monte Carlo simulation
- Samsung Research, Data Analytics Lab (2019) — Multilingual NLP for semantic extraction from B2C communications
- Kyutech Institute of Technology, Japan (2018) — Elderly care robotics with Mask R-CNN + Baxter robot. Featured on Japanese National TV.
Publications:
- APEX: Adaptive Ext4 File System for Enhanced Data Recoverability — IEEE Cloud Computing 2019
- Transformative Effects of IoT, Blockchain and AI on Cloud Computing — cs.DC 2019
| KVPY Fellowship | All India Rank 3 / 500,000 students (IISc Bangalore, 2015) |
| JEE Advanced | All India Rank 81 / 200,000 qualified (2016) |
| IIT Delhi | B.Tech CS, CGPA 9.10/10.0, Institute Merit Award (top 7%) |
| Samsung | Best Paper Award (2022), AI & Cloud certifications, Data Science L2 |
| Physics Olympiad | National Top 35, Gold Medal (HBCSE-Bombay) |
Building toward the intersection of agentic AI systems and real-world deployment at scale. Interested in the full stack of post-training — from reward signal design to evaluation infrastructure to production serving. Long-term: building something of my own.
