Join Microsoft Research India as a Senior Researcher focused on Self-Evolving Agentic Systems. This role is at the forefront of AI innovation, advancing how autonomous agents learn and improve over time. You'll contribute to systems that go beyond static prompts, developing agents capable of continuous self-improvement, learning from failures, and adapting in real-time.

Our team is pioneering the development of agents that enhance their own policies, skills, memory, and underlying structures through interaction and reinforcement learning. This is an exceptional opportunity to shape the future of intelligent agents.

As a Senior Researcher, you will spearhead advancements in reinforcement learning for LLM-based agents. Your work will span algorithm design, reward signal creation, environment development, and scaling research concepts into production-ready systems.

You will collaborate extensively with researchers and engineers globally within Microsoft Research and across product groups, ensuring your contributions impact both the academic community and millions of users worldwide.

Ideal candidates possess a strong passion for reinforcement learning, LLM post-training, and autonomous agents, supported by a robust research publication record in top-tier venues or significant open-source contributions. We seek individuals who blend scientific excellence with practical engineering skills to bring research to life.

We are seeking candidates with a PhD in Computer Science, Machine Learning, Mathematics, or a related field, alongside at least one year of professional software engineering or research experience. Expertise in reinforcement learning and Large Language Models (LLMs) is essential, coupled with a proven track record in machine learning or computer systems research.

Demonstrated ability to conduct independent research, communicate findings effectively through publications or prototypes, and collaborate across diverse teams is crucial. Preferred qualifications include experience building and scaling AI systems, applying RL to AI models and agents, and hands-on experience with various AI training techniques and dataset development, particularly for multimodal or agentic tasks.