Introduction
Companies seeking to develop artificial intelligence solutions for industry-specific challenges are increasingly turning to Oregon State University. The AI master’s capstone, which is now entering its third year, has become the front door. The nine-month projects are the experiential cornerstone of the nation’s first multidisciplinary graduate program offering both master’s and doctoral degrees in AI, backed by a school with more than 30 faculty members working in the field and a research tradition in artificial intelligence that stretches back four decades.
The structure is straightforward. Master’s students may complete a capstone project in lieu of a thesis, working in small teams across three 10-week terms under the guidance of Alan Fern, professor of computer science, who leads the capstone sequence and advises students throughout development. Partners contribute a real problem, a dedicated mentor, and data; students contribute state-of-the-art AI skills and sustained, focused effort. There are no required fees, and the university makes no claim on resulting intellectual property. These make the capstone one of the lowest-risk ways for an organization to explore what AI can do for its business.
Experience from the first few capstone cycles shows that a broad range of partnerships thrive. The organizations that get the most out of the program can be global companies like HP or Portland startups like STEELPORT Knife Co., but they all share three traits: an invested point person who meets with students regularly, a problem scoped realistically to nine months, and data that is prepared and labeled, or a clear plan to produce it. Partners who return year after year see compounding returns, as each cohort can build on the last.
professor of computer science
Blue Primary, Yellow Secondary
The 2026 showcase: From the factory floor to the patent office
This year’s cohort presented projects at the program’s annual showcase in June, demonstrating how AI can improve access to information, support decision-making, and streamline complex workflows.
Two projects served HP’s PageWide Industrial division, which manufactures industrial printing presses. In the first project, Caitlyn Lewis developed Front Desk, a natural-language interface that frees HP personnel from pre-built dashboard reports. By integrating large language models into the data-access workflow, users can ask questions in plain English and get relevant answers - with iterative prompt refinement delivering measurable gains in answer quality, SQL performance, and response latency.
In the second project, Yuqing Liu, Xiaoyan “Sean” Yang, and Shashi Kanth Koppala tackled a problem every press operator knows: configuring an industrial inkjet press means balancing dryer power, press speed, web tensions, moisture levels, and other interdependent settings. This is a process that has relied heavily on trial and error in the past. Their closed-loop recommendation system learns optimal configurations from accumulated operator feedback, evaluates candidate settings against similar historical jobs, and estimates expected print quality. A companion dashboard lets operators explore comparable runs and understand why the system recommends what it does.
Two more teams built AI agents for research and innovation at Oregon State itself. Carson Nitta, Woonki Kim, Aishwarya Dattatray Joshi, and Parnashri Nandam created the Grant Matchmaking Agent, which aggregates funding opportunities from multiple agencies and matches them to faculty using semantic embeddings, keyword extraction, and ranking algorithms applied to profiles, publications, and CVs. As faculty update their profiles, the recommendations adapt to evolving research directions.
Pranav Pravin Mane, Prayoga, and Pranav Vasist built InventionID, a web-based prototype that identifies potentially patentable ideas in technical papers. Because patent-search language differs from research-writing language, evaluating papers for patentable innovations is time-intensive. InventionID automatically extracts the innovation from an uploaded paper, generates patent searches, reviews related patents, and delivers an interactive report summarizing the invention and relevant prior art.
“Four years ago, most of these projects centered on building a model. This past year, we had teams that delivered complete systems, such as agents that search patents, interfaces that let you talk to your data, and recommenders that learn from operators on the factory floor. That shift mirrors exactly what's happening in industry, and our students get to experience it firsthand,” Fern said.
Join the next cohort
The School of Electrical Engineering and Computer Science is now soliciting industry and research partners for the coming year. Organizations interested in sponsoring or proposing a project can submit an interest form through the program’s online portal by September 16th.
Submit Proposals at M.S. AI Capstone Projects. For more information about the AI capstone projects, contact Alan Fern.
