Computer Science Students Make Strides in AI

“These projects are real-world projects with real-world implications,” said Mario Lim ’95, adjunct professor of computer science and cofounder of Openprise, a software startup in San Mateo.

Here, some of the students describe their projects.

Project: Cloud Pathology and AI
Team: Ananya Agarwal MS ’24, Haebin Lee MS ’25, Rucha Maslekar MS ’25
Supervisor:  Gaurav Kalele MS ’23, founder, Connection Loops, San Francisco

Problem: “When a patient gets a medical test such as a blood test, a lab doctor called a pathologist needs to review the test numbers and figure out what the numbers mean for each patient,” said Haebin Lee MS ’25. “It takes time for doctors to interpret the test results, and patients have to wait.”

Solution: “Cloud Pathology software manages laboratory data, including patient information and test results,” Lee said. “In seconds, our Cloud Pathology AI tool analyzes and summarizes a patient’s results on a blood test, for example. All a doctor has to do is to check the analysis and summary to make sure they’re correct, and then pass them to the patient. These summaries are easy to understand and give patients tips for staying healthy. Patients get their test results sooner and can start following the health advice right away.”
 

Project: GeoAI, Automating Environmental Detection
Team: Hana Gamracy ’24, Noga Gottlieb ’24, Paramdeep Sodhi ’25, George Sphicas ’25
Supervisors: Professors David Guy Brizan and Chris Brooks

Problem: “David Saah, professor of environmental science, asked us a semi-big question at the GenAI summit at USF last April,” said Paramdeep Sodhi ’25. “Could we create a geoanalysis chatbot that allows users to quickly locate and interrogate geospatial data?”

Solution: “In CS 490 Senior Team class last fall, we did a capstone project called GeoAI, which is a chatbot designed for geospatial analysis,” Sodhi said. “We took a geospatial analysis tool called Clay and trained it to use geospatial data, such as satellite images, to detect things. As an exercise, we trained Clay to detect backyard swimming pools in the Bay Area and Los Angeles County. From there, we advanced to tackling real-world challenges like identifying burn scars and flood plains. Others could use Clay to detect illegal logging roads in the Brazilian rainforest or unauthorized mining roads in Ghana. It’s all about collecting and analyzing data to better manage our planet’s resources. With instructions from just a few humans, the GeoAI chatbot uses Clay to analyze more geospatial data than a hundred humans could ever handle.”

Project: SnapLogic Agent Framework
Team: Yin Hu ’25, Qianru Wei MS ’24, Heran Zhang MS ’25 
Supervisor: Greg Benson, professor at USF and chief scientist at SnapLogic, San Mateo

Problem: “The original SnapLogic GenAI App Builder enables users to send questions to large language models — such as OpenAI GPT, Google Gemini, or Anthropic Claude — and receive responses. This is great, but the GenAI App Builder doesn’t enable users to create their own agents from scratch,” said Heran Zhang MS ’25.

Solution: “SnapLogic Agent Framework is the next-generation app builder,” Zhang said. “The agent framework enables users to create agents. Think of agents as complex prompts or questions or tasks that users can create and assign to large language models. SnapLogic Agent Framework is a part of SnapLogic AgentCreator, and SnapLogic AgentCreator enables users to create AI-powered ‘assistants.’ An AI assistant can handle tasks such as automating processes, making decisions, and managing data. For example, an assistant can analyze contracts, legal documents, or policies, flagging inconsistencies or potential compliance issues. It simplifies regulatory research by summarizing relevant laws, reducing the time and cost of legal reviews for businesses.”

Project: REST 8 Service
Team: Yahvi Bhatnagar ’24, Andrew Diep ’26, Colby Dobson MS ’26, Shreyash Hamal ’25, Johanna Lazaro ’25
Supervisor: Consilient Labs (Holly Potter ’23, Angela Rubalcava ’23, Reiner Steffens), San Francisco

Problem: “On software development projects, many workers work together on a computer server, but not all workers are allowed to access the entire server or perform all actions,” said Yahvi Bhatnagar ’24. “If they go where they’re not allowed to go or make changes they’re not authorized to make, they can damage the software or delay the project.”  

Solution: “REST 8 is a tool that controls access to servers. Think of it as a gatekeeper,” Bhatnagar said. “Each authorized user is given an encrypted token or key that grants them access, but they can only access the areas for which they have permission. It’s similar to a Google document where the owner assigns access privileges. In this case, users can be designated as viewers (look around but don’t make changes), members (work in specific areas), or administrators (access all areas). REST 8 helps software teams maintain control over their projects while protecting privacy and ensuring security.”