Token Foundry-Research Scientist / Research Engineer-Qwen Research

Role Overview

We are exploring foundation model systems that can keep learning from environments and their own experience. You will work on recursive self-improvement, reinforcement learning, and coding agents.

What You Will Do

1. Develop methods that enable models to plan, experiment, learn from failures, and improve over time.

2. Explore how models and agent harnesses can co-evolve through interaction and feedback.

3. Develop research agents and multi-agent AI systems that support MLE, post-training, and scientific workflows.

4. Advance reinforcement learning for long-horizon tasks and sustained capability growth.

5. Develop robust rewards and verifiers for complex, open-ended tasks, enabling effective learning from long-horizon agent interactions.

1. Strong foundations in machine learning and deep learning, with experience in LLM training, reinforcement learning, and agents.

2. Ability to turn ambiguous research questions into rigorous experiments and working systems.

3. Strong programming, communication, and collaboration skills.

4. Preferred Qualifications:

(1) Experience in LLM reinforcement learning, automated AI research, or long horizon agents.

(2) Contributions to end to end projects spanning data, training, evaluation, and systems.