Algorithm Researcher — Agentic Vision
Core
Lead full-stack algorithm R&D for post-training of Vision-Language Models (VLMs), optimizing visual perception, reducing hallucinations, and enhancing user interaction scenarios.
Role type
Senior IC algorithm researcher (agentic vision & VLM post-training)
Builds
Vision foundation models with advanced visual perception, reasoning, and execution capabilities for real-world user scenarios
Domain
Artificial Intelligence / Large Language Models / Computer Vision
Deliverable
production ML models
Required skills
Supervised Fine-Tuning (SFT), Reinforcement Learning (RLHF, PPO, GRPO), Transformer architecture, Visual Agent solutions, Hallucination detection & removal, Distributed training, Large-scale model training (tens/hundreds of billions parameters), Python, PyTorch, TensorFlow, Megatron-LM, DeepSpeed, vLLM
Preferred skills
User profiling, Recommendation systems, Long-context/long-term memory optimization, Top-tier conference publications (NeurIPS, ICLR, ICML, ACL, EMNLP), High-impact open-source contributions
Technologies
VLMs, RLHF, PPO, GRPO, Megatron-LM, DeepSpeed, vLLM, PyTorch, TensorFlow
Responsibilities
Lead post-training R&D for Vision-Language Models; Optimize low-level visual perception and spatial awareness; Reduce hallucinations in multimodal models; Align models for real-world user interaction scenarios
Seniority
Senior, hands-on IC
