Algorithm Research Intern — Agentic Vision 108335
Core
Research and development on Agentic Vision for Hunyuan multimodal foundation models, focusing on visual reasoning post-training and spatial intelligence.
Role type
Research intern (Agentic Vision / Multimodal Foundation Models)
Builds
Multimodal post-training methods, verifiable data synthesis pipelines, and interactive sandboxes for visual reasoning agents.
Domain
Artificial Intelligence / Computer Vision / Multimodal Foundation Models
Deliverable
production ML models
Required skills
Deep learning, algorithm implementation, PyTorch, distributed training, visual reasoning, reinforcement learning (PPO/GRPO), reward modeling, system-level design, experimental analysis
Preferred skills
Experience with verl, OpenRLHF, Megatron, RLVR, data mixture optimization, embodied spatial planning
Technologies
PyTorch, PPO, GRPO, OPD, verl, OpenRLHF, Megatron, rendering engines, code sandboxes
Responsibilities
Conduct R&D on visual reasoning post-training; develop multimodal post-training methods; build verifiable data synthesis pipelines; build interactive sandboxes; develop evaluation and error attribution methods.
Seniority
Intern
