Research Scientist - Reinforcement Learning
Institute of Foundation Models · Sunnyvale, CA ·
- Work mode
- Onsite
- Experience
- 3+ years
Institute of Foundation Models · Sunnyvale, CA ·
About the Institute of Foundation Models
We are a dedicated research lab for building, understanding, using, and risk-managing foundation models. Our mandate is to advance research, nurture the next generation of AI builders, and drive transformative contributions to a knowledge-driven economy.
As part of our team, you’ll have the opportunity to work on the core of cutting-edge foundation model training, alongside world-class researchers, data scientists, and engineers, tackling the most fundamental and impactful challenges in AI development. You will participate in the development of groundbreaking AI solutions that have the potential to reshape entire industries. Strategic and innovative problem-solving skills will be instrumental in establishing MBZUAI as a global hub for high-performance computing in deep learning, driving impactful discoveries that inspire the next generation of AI pioneers.
Position Summary
As a Research Scientist within our Reinforcement Learning team, you will play a fundamental role in establishing our scientific and technical directions toward the development of emergent capabilities within Foundation Models. The role involves pioneering novel approaches within Reinforcement Learning to facilitate paradigm shifts in foundation modeling. The role involves prototyping and adapting novel approaches to learning from experience, contributing to large-scale RL training infrastructure, and produce replicable code for public release. You will also be expected to build and maintain a productive research portfolio, supported by internal and external collaborations.
Develop novel research toward massive scale self-play for foundation model training, agentic tasks, and imbuing models with the capability to proactively learn from its environment.
Initiate and pursue novel reinforcement learning algorithmic approaches to define and drive emergent capabilities in Foundation Models.
Full-stack engineering from data curation, model architecture and algorithm design, to final production of models for end-users using high quality (documented, tested, maintainable) code.
Contribute to technical reports and research publications.
Represent MBZUAI at industry conferences and events, showcasing the institution’s technology and deep learning capabilities and establishing MBZUAI as a global leader in AI research and innovation.
Proactively engage with the open-source community.
Contribute to large-scale reinforcement learning training and inference frameworks.
Facilitate internal and external collaboration
Minimum
3+ years of hands-on experience with reinforcement learning
Demonstrated ability to independently identify limitations of current practice (internal and external), formulate and enact solution strategies for improvement.
Proactive mindset with the ability to identify impactful research questions and execute on them with minimal supervision.
Strong Python development skills with a focus on research-grade code and scalable data pipelines.
Practical experience implementing complex mathematical concepts into reliable, well-documented code.
Experience applying novel RL algorithms to practical applications.
Strong experience contributing to academic and/or open-source research through publication, GitHub contributions, or professional presentations.
Strong communication and collaboration skills for effective cross-functional work.
Strong systems and engineering expertise in deep learning frameworks such as PyTorch, Jax, etc.
Experience in large-scale model training (LLMs or Diffusion Models) on large clusters.
Familiarity with current RL+LLM training libraries
Experience training policies in self-play, possibly demonstrated by publication, blog post, public code.
Experience working with Diffusion Models in RL, possibly demonstrated by publication, blog post, public code.
Strong publication record in leading AI and RL venues (e.g.ICLR, ICML, NeurIPS, RLC, JMLR, TMLR)
Familiarity with performance constraints in production environments and the trade-offs in model design and execution.
Prior contributions to open-source ML research or data tools.
Demonstrated ability to solve complex system-level challenges and debug failures across training/inference stack (e.g. memory issues, deadlocks, I/O bottlenecks, multi-node communication failures).