Multi-Modality Research Intern

Tencent International Service Pte. Ltd.; Singapore; On-site; SGD 90,000 - 130,000; Full time

Job summary

Tencent International Service Pte. Ltd. in Singapore seeks a research scientist to pioneer multimodal understanding, post-training methods, and agent planning in real gaming scenarios.

You will design end-to-end experiments and collaborate with engineers to deploy prototypes at scale. You will advance multimodal model capabilities across images, video, text, and audio, publish at top conferences, and contribute solutions for diverse in-game applications while grounding research in production

Responsibilities

• Pioneer new research directions in multimodal understanding, post-training, reasoning, grounding, and agent planning, grounded in real gaming scenarios.
• Abstract research problems from game business scenarios, solve them, and land solutions that serve diverse in-game applications.
• Advance the understanding capabilities of multimodal large models (VLMs/MLLMs) across images, video, text, and audio.
• Lead post-training methodologies — SFT, RLHF/RLAIF, reward modeling, and preference alignment — to improve model capability and reliability.
• Design and execute experiments end-to-end: from data and benchmarking, to training, evaluation, and iteration.
• Collaborate with engineers to bring research prototypes into production and deploy them across game scenarios at scale.
• Publish at top-tier conferences and contribute to the broader research community.

Job description

• Pioneer new research directions in multimodal understanding, post-training, reasoning, grounding, and agent planning, grounded in real gaming scenarios.
• Abstract research problems from game business scenarios, solve them, and land solutions that serve diverse in-game applications.
• Advance the understanding capabilities of multimodal large models (VLMs/MLLMs) across images, video, text, and audio.
• Lead post-training methodologies — SFT, RLHF/RLAIF, reward modeling, and preference alignment — to improve model capability and reliability.
• Design and execute experiments end-to-end: from data and benchmarking, to training, evaluation, and iteration.
• Collaborate with engineers to bring research prototypes into production and deploy them across game scenarios at scale.
• Publish at top-tier conferences and contribute to the broader research community.