Algorithm Research Intern - Agentic Vision

About the role

Tencent's Technology Engineering Group (TEG) is recruiting an Algorithm Research Intern focused on Agentic Vision. The intern will research and develop Agentic Vision for Hunyuan multimodal foundation models, with a focus on visual reasoning post-training, within the TEG team that supports technology and operational platforms and runs the largest networking, device and data centre network in Asia.

Key responsibilities

• Research and develop Agentic Vision for Hunyuan multimodal models, integrating visual perception, planning, tool and code execution, environment feedback and self-correction for dynamic, interactive visual reasoning.
• Advance visual reasoning and spatial intelligence, including fine-grained perception, diagram reasoning, 3D spatial understanding, complex visual logic and embodied spatial planning.
• Develop multimodal post-training methods such as cold-start SFT, online reinforcement learning (PPO / GRPO), on-policy distillation, and verifiers and reward models, addressing credit assignment and long-horizon context.
• Build verifiable data synthesis pipelines using rendering engines, symbolic systems and code sandboxes, and establish a continuously improving data flywheel.
• Build interactive sandboxes for code execution, visual tools and search, and develop evaluation and error attribution methods across perception accuracy, reasoning correctness, tool efficiency and task completion.

Required profile

• Master's degree or higher in Computer Science, AI, Computer Vision, Machine Learning, Mathematical Statistics or a related field.
• Experience in multimodal foundation models, visual reasoning, multimodal agents or reinforcement learning for post-training.
• Familiarity with post-training methods and hands-on experience in one or more of SFT, PPO / GRPO, OPD, RLVR or reward modeling.
• Strong system-level design and end-to-end execution skills, with independent technical judgment and the ability to turn ideas into validated experiments.

Required skills

• Strong foundation in deep learning and algorithm implementation, with PyTorch proficiency and distributed training experience.
• Experience with verl, OpenRLHF, Megatron or similar frameworks is a plus.