Roberts Innovation Fund to Support Inventions in AI, Environmental Technology, and Healthcare
New methods for removing carbon from seawater, personalized heart surgery simulations, and ultra-low voltage circuits for energy-efficient AI are among the groundbreaking projects funded by the 2025 Roberts Innovation Fund Awards. This year, the Fund provides $500K in accelerator funding to support 10 new inventions led by faculty from Yale's School of Engineering & Applied Science.
The Roberts Innovation Fund, an accelerator fund created in 2022, managed by Yale Ventures and overseen by Yale Engineering's Office of the Dean, focuses on supporting technologies with significant potential to benefit a wide-ranging number of fields. In addition to funding, awardees receive world-class support and mentorship with access to industry experts and more than $1.8M in cloud computing credits and other resources from Amazon, Google, and Microsoft. The awardees will present their breakthrough projects for the first time at the Yale Innovation Summit on May 28, 2025.
In line with Yale Engineering's strategic vision, the Roberts Innovation Fund bolsters the School's culture of innovation and entrepreneurship by advancing an ecosystem that brings to market discoveries that will benefit humanity. Yale Engineering Dean Jeffrey Brock noted that the significant number of outstanding proposals this year underscores both the broad range of technologies the Roberts Innovation Fund can support and the strong and growing interest from Yale Engineering faculty in these opportunities. This year applicants hailed from nearly all Yale Engineering departments, including computer science, biomedical engineering, mechanical engineering, materials science, applied physics, electrical and computer engineering, and chemical and environmental engineering.
"The Roberts Innovation Fund showcases Yale Engineering's commitment to addressing global challenges through innovation, and the diversity of this year's winning projects – from energy-efficient AI to new seawater carbon-capture techniques – reflects the vast range of those challenges," said Yale Engineering Dean Jeffrey Brock. "With awardees from nearly every engineering department, we're seeing our strategic vision in action. To make a real-world impact, these groundbreaking faculty projects must make the leap from laboratory research to market-ready technologies, and the Roberts Fund helps to bridge that gap."
The Roberts Fund continues to evolve in its third year, broadening its scope to address a wider range of technological challenges while strengthening connections between academic innovation and industry applications.
"This year, we were once again reminded of the potential that Yale Engineering projects have to impact some of the world's most pressing challenges," said the Fund's director, Claudia Reuter. "It's a privilege to be able to support teams working on initiatives ranging from advances in AI to water quality issues."
The Roberts Innovation Fund is open to Yale faculty with a primary or secondary appointment in the School of Engineering & Applied Science with a novel innovation that solves a significant problem with the potential for scale. Applicants must demonstrate a proof of concept and a clear articulation for the use of the awarded funds.
One of the awardees, Rex Ying, assistant professor of computer science, said he's looking forward to working with the Roberts Fund on his project, with Yale researcher Ali Maatouk, to develop an automated domain-expert LLM for any domain.
"The Roberts Innovation Fund provides exactly the kind of support we need to bridge the gap between academic research and real-world deployment for our specialized LLM technology," Ying said. "With this funding, we aim to democratize access to domain-specific AI expertise, enabling users to create custom AI assistants tailored to specialized fields from medicine to engineering with just a simple request. The mentorship and resources that come with the Roberts award will be invaluable as we work to bring this technology to market and help organizations leverage the power of specialized AI without requiring deep technical expertise."
This year's awardees are:
Canary Sensor for household water purification systems
- Jaehong Kim, Henry P. Becton Sr. Professor of Chemical & Environmental Engineering
- Rilyn Todd, Ph.D. candidate
- Yonghyeon Kim, Postdoctoral Associate
The global market for household water purification systems is valued at $35B and continues to grow. Current systems lack sensors to monitor the quality of treated water. When these systems fail, harmful pollutants can be released and impact health. Their technology aims to help users know when the treated water contains pollutants. Just as miners in the past used canaries as air pollution sensors, by knowing to evacuate if a canary became ill, the team's novel Canary Sensor will provide a similar warning for water quality.
Carvis Health
- James S. Duncan, Ebenezer K. Hunt Professor of Biomedical Engineering, Radiology & Biomedical Imaging, Electrical & Computer Engineering and Statistics & Data Science
- Daniel Pak, Postdoctoral Fellow
- Theodore Kim, Professor of Computer Science
The team has developed a software solution to provide scalable personalized simulations of heart surgeries, with the goal of decreasing complexity in surgical planning, and increasing the efficacy of the procedure. They intend to improve patient care, reduce complications by 50%, saving $2.9B in complication costs and decrease burn-out with minimal changes to overall workflow.
Compact ultra-high stability vacuum-gap reference cavities enabling portable precision metrology and sensing
- Peter Rakich, Donna L. Dubinsky Professor of Applied Physics & Physics
- Greg Luther, PhD, CEO and co-Founder of Resonance Micro Technologies Inc.
A wide range of technologies, including navigation (e.g. GPS), radar, sensing, and communications technologies rely on low-noise oscillators (i.e. clocks). The lower the noise of the oscillator, the more accurately users can identify the position of a vehicle, detect a microwave communications signal, or sense motion in a radar system. The Rakich Lab has invented first-of-a-kind ultra-high performance micro-resonators and oscillators that fill critical technology gaps in both classical and quantum application spaces.
Q0 AI: The Behavioral Intelligence Layer for AI Agents
- David van Dijk, Assistant Professor of Medicine and Computer Science
- Olin Geimer, Postgraduate Fellow
- Ivan Vrkic, Postgraduate Associate
Q0 AI provides the intelligence layer for a digital workforce—teams of AI agents—that collect, manage, and analyze human-generated data. Every organization depends on human behavior, making a deep understanding of these patterns essential for success. Our advanced behavioral foundation models and intelligent analytics tools seamlessly integrate into solutions powered by teams of AI agents, enabling organizations across business, policy, marketing, and healthcare to optimize services, drive value, and effectively leverage the core behavioral patterns.
ExpBuddy: Domain Expertise Agent in the Cloud via Specialized Large Language Models
- Rex Ying, Assistant Professor of Computer Science
- Ali Maatouk, Postdoctoral Associate
Despite their success, Large Language Models (LLMs) lack specialization, as they focus on knowledge from diverse domains. This lack of specialization hinders their performance compared to domain-specific LLMs. The team is working to create a unified cloud service to develop specialized LLMs for any domain of interest through a simple interaction with a web-client interface. Leveraging their AI product, users can create domain-expert LLMs for any field, no matter how specific, through a process that is transparent and accessible to the user.
Low-Cost Soil Sensing with Wi-Fi and Machine Learning
- Leandros Tassiulas, John C. Malone Professor of Electrical & Computer Engineering & Computer Science
- Jian Ding, Ph.D. candidate
Accurate, inexpensive, and rapid determination of soil properties is an essential requirement for data-driven agriculture. In this project, the team has focused on using widely accessible Wi-Fi hardware and smartphone images to measure soil moisture, electrical conductivity, and carbon content—three essential properties for monitoring soil health and mitigating climate change.
Millivolt AI: Ultra-low voltage circuits for energy-efficient AI
- Logan Wright, Assistant Professor of Applied Physics
- Jinchen Zhao, Ph.D. candidate
- Jared Wyetzner, Yale College '26 (Physics)
Millivolt AI aims to revolutionize AI hardware by enabling inference and training to use up to a million times less energy. Practically, their technology is in many ways a software upgrade that relies on deep insight into hardware physics. Their vision is to power large-scale sustainable AI-development.
PARTEE: Rich I/O and availability for safety-critical, privacy-sensitive robotics and IoT software
- Zhong Shao, Thomas L. Kempner Professor of Computer Science
- Richard Habeeb, Ph.D. candidate
PARTEE aims to protect the most safety-critical software tasks running on modern robotics while also providing access to complex, rich I/O. The team is working to combine a few major technologies: a partitioning system which isolates critical software, a secure inter-task communication protocol, and a mechanism to send and receive complex I/O.
Photochemical Carbon Removal from Seawater
- David Kwabi, Associate Professor of Chemical & Environmental Engineering
- Byron Ross, Incoming Postdoctoral Fellow
- Bin Yun, Ph.D. candidate
- Ava Canney, Yale College'26 (Chemical Engineering)
The Kwabi Lab has pioneered a groundbreaking approach to carbon mitigation by leveraging the ocean's natural ability to absorb CO₂ from the atmosphere. Traditional photochemical and electrochemical systems for seawater carbon removal are often complex and energy-intensive, limiting their scalability and widespread adoption. By harnessing light-driven carbon removal, the team is reimagining carbon capture at scale, offering a next-generation solution to one of the world's most pressing climate challenges.
SciMentor: LLM Agents as Research Advisors
- Arman Cohan, Assistant Professor of Computer Science
- Yilun Zhao, Ph.D. candidate
- Alan Li, Ph.D. candidate
SciMentor is an AI-powered platform that serves as a virtual research advisor for early-stage researchers. SciMentor leverages advanced Large Language Models (LLMs) within an innovative LLM-Agent framework to provide adaptive, iterative mentorship, mirroring the guidance of domain-specific experts. It addresses key challenges faced by junior researchers, including limited access to mentorship, time-consuming literature reviews, and the need for rapid hypothesis validation.