Three Carnegie Mellon Faculty Named 2026 AI2050 Early Career Fellows

The fellowship and the cohort

Akari Asai and Daniel Fried, both assistant professors in the Language Technologies Institute, and Andrea Bajcsy, an assistant professor in the Robotics Institute, are among 22 early career researchers selected for this year's cohort.

According to Carnegie Mellon, Schmidt Sciences selected 28 researchers overall, including six senior fellows. The organization will provide the cohort with up to $16 million over three years to pursue their research.

AI2050 was launched in 2022. It supports researchers working on opportunities and challenges that could shape how AI benefits society in the coming decades. The program focuses on a working list of hard problems in AI, including ensuring AI systems are safe and reliable and using the technology to address major scientific and societal challenges.

What the three researchers will work on

Akari Asai develops reliable and adaptable language systems, particularly retrieval-augmented and agentic language models. Her fellowship-supported research will explore how AI agents can better support scientific discovery. That includes using scientific literature, data and tools, interacting with laboratory environments, and evaluating scientific claims based on evidence.

Andrea Bajcsy works at the intersection of robotics, machine learning and AI safety. Her research addresses a growing challenge: general-purpose robot "brains" trained on massive, hidden datasets that can behave unpredictably once they're deployed. She will develop safety steering methods to guide robot behavior, helping robots interpret instructions, predict the consequences of their actions and ask for help.

Daniel Fried leads a research group called Language Agents, Reasoning and Communication, which works on natural language processing and the interaction between people and AI systems, with a particular focus on large language model agents. His fellowship-supported research will explore trustworthy communication among teams of AI agents. Planned methods include allowing agents to verify commitments, communicate based on explicit plans and make those verification processes understandable to humans.

Support beyond funding

The 2026 fellows also receive opportunities to collaborate with other AI2050 researchers. They can attend an annual gathering to share their work and build connections across disciplines. According to the school, fellows may receive additional funding for collaborations and computing resources.

The three projects cover different areas: AI agents for science, safer general-purpose robots and trustworthy communication between AI agents. Each is tied to the reliability and safety themes that AI2050 names as priorities. Further details about the program are available on the Schmidt Sciences website.