LLM Agent Research Intern
LIGHTSPEED (OA Huoshui Platform) Open · verified Oct 22, 2026 At a glance
Junior AI Engineering role at LIGHTSPEED (OA Huoshui Platform). United Kingdom-London · full-time.
Pay not stated AI Engineering salary
aimlroles summary, based on the employer's posting.
What you'll do
- Train in-character game agents with supervised fine-tuning, then apply preference optimization and verifiable-reward reinforcement learning
- Prepare dialogue datasets for training and preference comparisons, covering memory, relationship stages, language changes and emotion
- Create automatic checks for format, language, persona limits and safety refusals, then use them in training rewards and regression tests
- Bring the agent into a live game with designers and engineers while balancing response speed, operating cost and player experience
What you bring
- You are enrolled in a relevant master's or PhD program, or recently graduated, and can work at least three days weekly
- You have hands-on LLM fine-tuning and post-training experience, including DPO and verifiable-reward reinforcement learning or a related method
- You have trained or assessed character-dialogue models for persona consistency and coherent multi-turn conversations
- You are proficient in Python and PyTorch and can use the Hugging Face Transformers training tools
Who this fits
This internship suits a graduate researcher who can run experiments from data preparation through evaluation and explain results and trade-offs clearly. Fluent written and spoken English is required, along with availability for at least three days a week; university and supervisor approval may be needed.
From the employer
About the Hiring Team
Level Infinite is Tencent’s global gaming brand. It is a global game publisher offering a comprehensive network of services for games, development teams, and studios around the world.We are dedicated to delivering engaging and original gaming experiences to a worldwide audience, whenever and wherever they choose to play while building a community that fosters inclusivity, connection, and accessibility. Level Infinite also provides a wide range of services and resources to our network of developers and partner studios around the world to help them unlock the true potential of their games.
What the Role Entails
We are hiring a research intern to train LLM agents that play in-game characters. The agent has to stay in persona across multi-turn dialogue, use memory and game context, and keep relationship, tone, and world rules intact. The work sits on a live role-play NPC model: supervised fine-tuning today, with preference optimization and verifiable-reward reinforcement learning as the next training stages.1. Train role-play agents with supervised fine-tuning (SFT), then post-train them with direct preference optimization (DPO) and reinforcement learning with verifiable rewards (RLVR). Iterate on persona consistency, multi-turn coherence, and instruction following.
2. Build and curate SFT, preference, and verifiable-reward data for character dialogue, covering memory injection, relationship stage, language switching, emotion, and out-of-character resistance.
3. Design checks that can be verified automatically, such as format, language, persona boundaries, and safety refusals, and use them both as RLVR rewards and as offline regression tests.
4. Work with game designers and engineers to bring the agent into a live game, and keep latency, cost, and player experience within production constraints.
Who We Look For
1. Currently enrolled in a Master's or PhD program in Computer Science, NLP, Machine Learning, or a related field, or recently graduated. Available at least three days per week, with supervisor and university approval where required.
2. Fluent English (written and spoken) and strong communication skills are required. You will work closely with game designers, engineers, and researchers across regions, and must be able to explain experiment results and trade-offs clearly.
3. Hands-on experience with supervised fine-tuning (SFT) of LLMs, and with post-training methods including DPO and RLVR (or a closely related verifiable-reward RL method). You should be able to run the loop from data and training config through evaluation and the next iteration.
4. Experience training or evaluating role-play or character-dialogue models, including persona consistency, multi-turn dialogue, and staying in character.
5. Proficient in Python and PyTorch, and comfortable with the Hugging Face / Transformers training stack.
6. Preferred: experience with LLM agents (memory, planning, or tool use), LLM applications in games, or a publication at a major NLP/ML venue (ACL, EMNLP, NAACL, NeurIPS, ICLR, ICML).
Equal Employment Opportunity at Tencent
As an equal opportunity employer, we firmly believe that diverse voices fuel our innovation and allow us to better serve our users and the community. We foster an environment where every employee of Tencent feels supported and inspired to achieve individual and common goals.