A year ago, the University of Pittsburgh Cloud Innovation Center (CIC), powered by AWS, opened its doors with a straightforward but ambitious mission: bring together Pitt students and cutting-edge cloud technology to solve real problems for the university community.
One year later, the results speak for themselves. Six open-source AI solutions were built, making an impact across health sciences and athletics. But the story of the Pitt CIC's first year isn't just about what was built. It's about who built it, and what building it meant for them.
What the CIC Experience Builds
While the CIC student interns gain plenty of hands-on tech skills, the true learning experience goes much deeper. Students at the CIC work with problems that go beyond the current technology landscape — they’re experiencing what it’s like to meet real stakeholder expectations and solve real problems. Those are skills that won’t go away even when technology changes.
They learn to ask the right questions. They learn to translate between what a researcher or coach wants and what a system can actually do. They learn to work with ambiguity, change direction when needed, and still ship something that works.
These are the skills that will define the workforce of the next decade — especially as AI tools become more capable and the competitive edge shifts toward people who know how to apply them thoughtfully.
🎓 The Students Behind the Solutions
At the heart of the Pitt CIC is a team of student developers who work alongside technical mentors and faculty partners to take ideas from concept to working prototype. These projects are about finding solutions that address real challenges for real stakeholders.
"This project broadened my horizons to see that every field has needs for computer and data science. Healthcare wasn't on my radar before, but now I see that any industry dealing with data has problems I can help solve," said Gary Farrell, Student Developer, Pitt CIC and Computer & Data Science double major, reflecting on his work building an AI-powered medical code mapping tool.
That kind of perspective shift from "I'm a developer" to "I'm a problem-solver who can work in any domain" is exactly what the CIC is designed to create. In a workforce that is changing faster than anyone can predict, the ability to apply technical skills to unfamiliar problems is one of the most valuable things a student can walk away with.
Over the course of the CIC's first year, student developers have built production-quality solutions using AWS Bedrock, Lambda, DynamoDB, SageMaker, and more. They've presented their work to stakeholders, navigated ambiguous real-world requirements, and shipped code that is now publicly available for anyone in the world to build on.
Six Solutions, Real Impact
In the last year, the CIC completed six projects. This by itself is worth celebrating, but the projects deserve some love, too. So, here’s how the CIC’s student developers found solutions for problems from across the University in just 12 months.
🏊 Automatic Highlight Reel Generator
Partners: Pitt Diving, Head Coach Katie Kasprzak
Coach Katie Kasprzak was spending roughly ten hours every week manually reviewing practice footage to find and evaluate individual dives. It was effective, but it was also a huge drain on time.
CIC student interns Roman Koshovnyk and Rowan Morse built the Automatic Highlight Reel Generator: a system that processes hours of practice footage, detects each individual dive, and automatically clips them into individual segments. After pivoting to a zero-shot vision-language model (VLM) to handle the real-world challenges of single-angle footage and water reflections, the solution achieves 97% dive detection accuracy.
"It transforms hours of manual review into just minutes," said Koshovnyk.
Coach Kasprzak put it simply: "This solution has completely transformed how we analyze practice, turning it from a massive drain on my time into a powerful tool that provides valuable quantifiable data that tracks the athletes' progress in a meaningful and user-friendly manner."
🏅 Diving Analytics Platform
Partners: Pitt Diving, Head Coach Katie Kasprzak
The same CIC partnership with Pitt Diving yielded a second solution: a platform to replace the manual, error-prone process of entering handwritten dive scores into spreadsheets.
Vincent Niedermayer and Varun Shelke built the Diving Analytics Platform, a web application that lets coaches upload photos of handwritten dive logs, which are then automatically parsed using OCR and a large language model, converted into structured data, and visualized in dynamic performance dashboards.
"We designed a solution that integrates into her existing workflow. She was already taking photos of each dive sheet after every session, then spending hours manually entering that data into spreadsheets," said Niedermayer. "Now she can upload all her photos at once, and within seconds the data populates each diver's profile, with graphs and KPIs that help her deliver optimal feedback to the team."
📣 Smart Outreach Hub
Partners: Pitt Athletics, Lee Roberts, Executive Associate AD for Philanthropy and Engagement
Pitt Athletics had thousands of potential supporters in their database. Their sales team, working manually, could only reach a fraction of them.
CIC student interns Mohammed Misran and Varun Shelke built the Smart Outreach Hub: an AI-powered SMS engagement system that automatically reaches out to leads, conducts natural conversations, and guides interested prospects toward ticket purchases or philanthropic contributions, escalating to a human sales rep when the moment is right.
"Working closely with Pitt Athletics gave us real insight into how sales teams operate and what they need," said Misran. "The interesting part was teaching the LLM to truly sound like a sales rep and know when to hand off to a human."
The system's serverless architecture scales automatically to handle thousands of concurrent conversations while keeping costs minimal.
📊 Survey Analysis Agent
Partners: Pitt Athletics
Pitt Athletics collects large volumes of fan feedback through surveys, but getting meaningful insights out of that data required navigating multiple tools and waiting on manual analysis cycles.
CIC student interns Gary Farrell and Varun Shelke built the Survey Analysis Agent: a conversational AI interface that lets Athletics staff ask natural-language questions directly about their survey data, surfacing patterns, trends, and sentiment insights on demand, without requiring any technical expertise to operate.
🔬 EHR Code Mapper
Partners: Pitt Health Sciences, Dr. Christopher Horvat, Associate Professor of Critical Care Medicine, Pediatrics, and Biomedical Informatics
Medical researchers spend hundreds of hours manually mapping proprietary hospital codes to standardized vocabularies like LOINC, SNOMED, and RxNorm, before they can even begin their actual research. As Dr. Horvat explained, something as simple as finding a patient's sodium level can require sorting through 329 different variables.
Gary Farrell built the EHR Code Mapper: a two-step AI pipeline that uses embedding-based semantic similarity to narrow the candidate pool, then deploys a large language model to evaluate and rank the most likely standard code matches. The solution outputs results following FHIR standards, the format required for seamless healthcare data exchange, and provides reasoning for each suggested mapping so researchers can validate quickly.
The result: 92% exact match accuracy and 98% top-3 accuracy, with the potential to reduce manual mapping time by 80% or more.
"This complexity is the single biggest barrier to meaningful use of EHR data," said Dr. Horvat. The EHR Code Mapper takes direct aim at that barrier and makes the solution freely available to researchers everywhere.
📁 PHI De-Identification Tool
Partners: Pitt Health Sciences, Dr. Gilles Clermont, Professor of Critical Care Medicine and Vice Chair for Research Operations
For Dr. Clermont's research team, de-identifying millions of clinical notes to comply with HIPAA was taking approximately one week per processing run, and every new study with different privacy requirements meant starting over from scratch.
Ava Luu and Mohammed Misran built the PHI De-Identification Tool: a configurable, scalable system that uses AWS Bedrock to dynamically identify all 18 categories of HIPAA-defined protected health information across clinical notes in various formats, supports both Safe Harbor and Limited Dataset configurations, and processes notes in parallel, reducing a week-long task to approximately four hours for roughly 10 million clinical notes.
"When working with healthcare data, privacy protection is a top priority. Our main focus was ensuring the solution was reliable with exhaustive testing and evaluation," said Luu.
The solution's human-in-the-loop review interface lets researchers see exactly what was flagged, make edits, and confirm before the final de-identification runs.
"This tool will fundamentally change our data preparation workflow," said Dr. Clermont. "This will significantly increase our research capacity by eliminating the long wait researchers currently face for each request."
Looking Ahead
The CIC's foundation, Pitt's health sciences expertise and athletics partnerships, remains as strong as ever. But the horizon is wider.
The Pitt CIC's second year brings an expanded vision: More students, more projects, and an expanded scope that goes beyond health sciences and athletics to serve any corner of the University community. Any problem where real challenges meet the potential of AI and cloud technology, the CIC can find a way to solve it.
Every project starts with a problem someone couldn't solve on their own. Every solution ends as open-source code on GitHub, freely available to anyone who needs it. And in between, a group of Pitt students show that It’s Possible at Pitt.
Want to bring your idea to the Pitt CIC? We accept project proposals from University of Pittsburgh staff and faculty. Submit your idea today.
The University of Pittsburgh Cloud Innovation Center, powered by AWS, builds impactful, scalable solutions using cloud computing, artificial intelligence, and machine learning. Learn more at digital.pitt.edu/cic.
