Research Intern - AI Frontiers - Reasoning & Agentic Models

Microsoft hybrid • Redmondintern

Research Internships at Microsoft provide a dynamic environment for research careers with a network of world-class research labs led by globally-recognized scientists and engineers, who pursue innovation in a range of scientific and technical disciplines to help solve complex challenges in diverse fields, including computing, healthcare, economics, and the environment.

The AI Frontier Lab at Microsoft Research is seeking candidates to advance the state of the art in agentic model capabilities -- creating models and agents that can reliably perform tasks across digital systems on behalf of humans,  combining automation, reasoning, and interaction capabilities to execute workflows end-to-end, leveraging both text-based environments (CLI tools, APIs, scripts, MCPs) and visual environments (GUI applications).

 

Our lab conducts cutting-edge research in AI and publishes findings in top-tier venue such as NeurIPS, ICLR, ICML, and others. We release models and libraries in the open-source (e.g. Phi-4, Phi-4-reaoning, AgentInstruct, AutoGen, MagenticOne) while also working within Microsoft's ecosystem to ship our AI technologies in multiple products.

 

We seek Research Interns with demonstrated ability for technical work and a proven record of influential publications on Artificial Intelligence.

 

Research areas of particular interest for this team include, but are not limited to:

  • Reinforcement learning approaches for improving logical and mathematical reasoning, tool use and computer use agents
  • Developing novel training algorithms for enhancing reasoning and action taking efficiency and reliability 
  • Exploring synthetic environment creation, multi-agent training and self-play for RL training
  • Exploring scaling laws between test-time and training-time compute
  • Advanced optimization techniques for efficient training of large-scale models

 

Our group takes a holistic approach to improving foundational models that includes a variety of data modalities (language, vision, multi-modal, and structured data) and modern model architectures.

Requirements

  • Reinforcement learning approaches for improving logical and mathematical reasoning, tool use and computer use agents
  • Developing novel training algorithms for enhancing reasoning and action taking efficiency and reliability
  • Exploring synthetic environment creation, multi-agent training and self-play for RL training
  • Exploring scaling laws between test-time and training-time compute
  • Advanced optimization techniques for efficient training of large-scale models