Job Description
The mission of the AI Frontiers lab is to expand the pareto frontier of Artificial Intelligence (AI) capabilities, efficiency, and safety through innovations in foundation models and learning agent platforms. Some of our projects include work on Small Language Models (e.g. Phi, Fara, Fara-multimodal, OmniParser) and agentic AI systems (e.g. AutoGen, MagenticOne, Magentic-UI). We are advancing the science of agents with research on Machine Learning (ML), Human-Computer Interaction (HCI) and AI. To bridge foundational breakthroughs with real-world agent autonomy, we are building an end-to-end agentic model stack that integrates pre-training, post-training, reinforcement learning (RL), multi-agent collaboration, agentic harness, and deployment. Our core objective is to create self-improving AI systems that can navigate dynamic enterprise environments by learning through experience and interaction. As a Senior Researcher – Agentic AI, you will play a crucial role in modeling agentic intelligence: developing agents that learn to self-correct, specialize and collectively improve through self-play as well as interaction. The role spans research on recursive self-improvement using self‑generated data, adaptive coordination and delegation across multiple models, continual learning without catastrophic forgetting, human-AI collaboration for long-horizon tasks, and autonomous skill acquisition driven by interaction‑identified gaps. This role encompasses the full research-to-deployment spectrum, requiring an active focus on building robust evaluation frameworks and high-fidelity synthetic data pipelines. You will also be instrumental in refining the underlying architectures to ensure agentic systems scale effectively in complex environments. Researchers in the AI Frontiers lab work in a vibrant, multidisciplinary environment with access to large‑scale pipelines and evaluation environments, the ability to ground research in real-world product scenarios, and an open publication policy designed to share breakthroughs in agentic autonomy with the global research community. Our current research focuses on the intersection of self-improving systems and multi-agent reasoning, specifically exploring how recursive improvement, adaptive coordination, and autonomous skill acquisition can enable agents to continuously evolve through interaction without losing prior capabilities.
