Principal Researcher - Agentic AI - Microsoft Research AI Frontiers

Microsoft • Redmond, WA

22h

About The Position

The AI Frontiers lab aims to push the boundaries of AI capabilities, efficiency, and safety through innovations in foundation models and learning agent platforms. Projects include work on Small Language Models (e.g., Phi, Fara) and agentic AI systems (e.g., AutoGen). The lab advances agent science through research in Machine Learning (ML), Human-Computer Interaction (HCI), and AI. To connect foundational breakthroughs with real-world agent autonomy, an end-to-end agentic model stack is being built, integrating pre-training, post-training, reinforcement learning (RL), multi-agent collaboration, agentic harness, and deployment. The core objective is to create self-improving AI systems capable of navigating dynamic enterprise environments by learning through experience and interaction.

Requirements

• Doctorate in Computer Science or relevant field AND 3+ years related research experience OR Master's Degree in Computer Science or relevant field AND 4+ years related research experience OR Bachelor's Degree in Computer Science or relevant field AND 6+ years related research experience OR equivalent experience.

Nice To Haves

• Experience with building and evaluating agentic models.
• Research record demonstrated by public artifacts like models, tools, code in AI space or publications at the following conferences: NeurIPS, ICML, ICLR, ACL, NAACL, COLM.
• Expertise in the technical scaling of models and a commitment to engineering excellence, including the development of generalized code and robust research infrastructure.
• Experience publishing academic papers as a lead author or essential contributor in the field of Artificial Intelligence, deep learning, natural language processing and/or reinforcement learning.
• Experience participating in a top conference in relevant research domain (i.e. organizing a workshop, hackathon, community engagement/relations).
• Demonstrable ability to define an ambitious, original research agenda.
• Ability to collaborate, communicate effectively, and technically lead multi-disciplinary team.
• Keen interest in real-world applications and impact.

Responsibilities

• Perform cutting-edge research in collaboration with other researchers, engineers, and product groups.
• Be part of research breakthroughs in the field.
• Realize ideas in products and services used worldwide.
• Model agentic intelligence: developing agents that learn to self-correct, specialize, and collectively improve through self-play and interaction.
• Conduct research on recursive self-improvement using self-generated data.
• Research adaptive coordination and delegation across multiple models.
• Research continual learning without catastrophic forgetting.
• Research human-AI collaboration for long-horizon tasks.
• Research autonomous skill acquisition driven by interaction-identified gaps.
• Build robust evaluation frameworks.
• Build high-fidelity synthetic data pipelines.
• Refine underlying architectures to ensure agentic systems scale effectively in complex environments.

Benefits

• Access to large scale pipelines and evaluation environments.
• Ability to ground research in real world product scenarios.
• Open publication policy designed to share breakthroughs in agentic autonomy with the global research community.
• Certain roles may be eligible for benefits and other compensation.

What This Job Offers

• Job Type: Full-time
• Career Level: Principal