From the lab to deployment, Carnegie Mellon University researchers build tools designed to make tangible differences in how people live, learn and work. That research recently earned top recognition in the Laude Institute(opens in new window)'s Moonshots competition, which concentrates on applying AI to some of society’s most pressing challenges, from workforce reskilling to expanding access to education.
The top award went to “ARISTOS: Reskilling for a Physical Workforce,” a pioneering CMU project that reimagines how hands-on skills are taught in an AI-driven world.
ARISTOS (ARtificial Intelligence for Successful Teaching Of Skills) addresses a critical gap in today’s AI landscape by focusing on physical, experience-based expertise that cannot be easily captured through text or data alone. The team is developing an AI-powered “virtual craft master” capable of guiding users through complex physical tasks in real time. Using advanced vision-language models, the system generates customized instructional videos and delivers adaptive, multimodal feedback tailored to each learner and their needs.
By making expert training available on demand, ARISTOS aims to fundamentally reshape how people acquire physical skills and remove barriers tied to cost, accessibility and location. The system is designed to reduce reskilling time compared to traditional approaches, whether through in-person instruction or static resources like online videos.
The hybrid training model combines simulated environments with real-world practice, allowing learners to begin training even without access to physical equipment and to build confidence in a safe, controlled setting before transitioning to hands-on work. The broader implications extend beyond individual learners. By lowering the cost and time required for training, ARISTOS could reduce industry-wide training expenses, accelerate workforce development, and open new opportunities for unemployed or undertrained workers.
"Dexterous physical work in unpredictable environments is one of the last things AI cannot simply automate, which makes it exactly where workforce reskilling efforts should be focused,” said Dave Patterson, founding board chair of the Laude Institute and chair of its Moonshots evaluation committee. “What this team is building — an AI system that generates customized instructional content and provides real-time guidance through complex physical tasks — creates a path to high-value skilled trades training that doesn't depend on geography or access to a master craftsperson."
The work is led by:
Changliu Liu(opens in new window), an associate professor in the CMU Robotics Institute(opens in new window) (RI).
Deva Ramanan(opens in new window), principal investigator and RI professor.
Raj Reddy(opens in new window), the Moza Bint Nasser University Professor in the RI.
Katia Sycara(opens in new window), the Edward Fredkin Research Professor in the RI.
Jun-Yan Zhu(opens in new window), the Michael B. Donohue Assistant Professor of Computer Science(opens in new window) and Robotics.
The ARISTOS team was one of eight winners to receive a $250,000 seed grant and a mandate to develop their ideas into fully scoped proposals for a $10 million Moonshot lab. Lab winners will be selected later this year.
Three CMU teams were also included among the competition's runners-up and honorable mentions. Runners-up will receive $200,000 and honorable mentions will receive $100,000, respectively. Later this year, Laude will host a dedicated showcase for these teams in front of an audience of funders.
Runner Up
“Build With, Not For: An AI-Accelerated Rapid Research Translation Platform for Equitable Reskilling”
- This project enables community-based organizations to build their own AI tools using a participatory platform. In partnership with Community Forge(opens in new window) in Pittsburgh, the team will co-design pilot tools to reduce administrative burden for social service workers.
- Team:
- Claire Le Goues(opens in new window), professor in the Software and Societal Systems Department (opens in new window)(S3D).
- Heather Miller(opens in new window), principal investigator and S3D assistant professor.
- Michael Skirpan(opens in new window), S3D assistant teaching professor.
Honorable Mentions
“A National Intelligence Infrastructure for AI-driven Workforce Transformation”
- This project develops a predictive system to track AI-driven workforce disruption that empowers individuals and organizations to identify and customize reskilling strategies.
- Team
- Ramayya Krishnan(opens in new window), the William W. and Ruth F. Cooper Professor of Management Science and Information Systems and dean emeritus of CMU's Heinz College of Information Systems and Public Policy(opens in new window).
- Rebecca Nugent(opens in new window), the Stephen E. and Joyce Fienberg Professor and head of the Department of Statistics and Data Science(opens in new window).
- Aarti Singh(opens in new window), the FORE Systems Professor in CMU's Machine Learning Department(opens in new window).
“Kaggle for Scientific Agents: AI-Driven Physical Experimentation”
- This project connects AI agents with automated labs to design and run physical experiments, specifically leveraging the CMU Biological and Chemical Innovation Cloud Lab as a unique hardware asset.
- Team
- Akari Asai(opens in new window), incoming assistant professor in the CMU Language Technologies Institute(opens in new window) (LTI).
- Lei Li(opens in new window), LTI associate professor.
- Graham Neubig(opens in new window), LTI associate professor.
- Newell Washburn(opens in new window), associate professor of chemistry and biomedical engineering in the Mellon College of Science(opens in new window).