Posts

The Computing Research Association (CRA) is a coalition of more than 200 organizations with the mission of enhancing innovation by joining with industry, government, and academia to strengthen research and advance education in computing. Every year, they recognize North American students who show phenomenal research potential with their Outstanding Undergraduate Researcher Award, and for 2024-2025, four Brown CS students received honors: Artem Agvanian and Corinn Tiffany (Finalists) alongside Byron Butaney and Kaleb Newman (Honorable Mentions).

Artem Agvanian

Artem participated in two research projects as an undergraduate researcher at Brown, both in the Efficient and Trustworthy Operating Systems (ETOS) group, advised by Brown CS faculty member Malte Schwarzkopf, designing private-by-construction systems that incur low performance and developer overhead to help application developers provide privacy guarantees to their users.

In their first project, called K9db, Artem built a compliant-by-construction relational database. In the second project, called Sesame, he designed an end-to-end privacy policy enforcement framework for web applications. Both projects were built towards reducing the risk of inadvertent privacy violations by developers and organizations.

“K9db aims to solve the problem of privacy compliance in relational databases by introducing an abstraction called a data ownership graph (DOG). Ownership-aware storage provides a mechanism for satisfying user data access and deletion requests for any applications that use the database,” Artem explains. “Sesame, by contrast, aims to solve the problem of accidental data leaks or misuse in web applications by enforcing the application’s privacy policies on the application’s code itself.”

“I am honored to be among the members of the Brown community and the ETOS group who have been elected to receive this award over the years,” Artem says. “It represents a recognition of our sustained efforts to make practical and impactful research systems.”

Corinn Tiffany

Corinn is a current senior working with Malte in the ETOS group, whose research focuses on privacy compliance and correctness in systems software. Her research projects include Sesame alongside Artem, and Sniffer, which is a static analysis tool to help developers find and audit “unsafe” code in their Rust code bases. Both Corinn and Artem are co-authors of the paper Sesame: Practical End-to-End Privacy Compliance with Policy Containers and Privacy Regions, which was published at the 2024 Symposium on Operating Systems Principles (SOSP).

“User data in web applications is governed by privacy policies, such as limits on the purposes for which data can be used and who the data can be revealed to. Without tooling like Sesame, developers must manually track what policy applies to what data, and attempt to apply the required checks when processing data or performing I/O,” Corinn says. “This is an error-prone process, and privacy bugs can lead to embarrassing data leaks and steep fines.”

Corinn then goes on to explain the Sniffer analysis tool, which targets auditing efforts for unsafe Rust code. Sniffer is now a collaboration between a team of researchers from both Brown and Stanford University. “System developers bootstrap guarantees like crash safety and data integrity from properties of safe Rust, but unsafe code can break them. Sniffer helps developers manage this risk by narrowing the scope of review efforts to unsafe that threatens a system’s guarantees,” Corinn says.

Corinn presented Sniffer at the SOSP 2024 Student Research Competition poster session and New England Systems Day, and a paper on this work is currently under submission. 

Byron Butaney

Byron’s research focuses on the 3D spatial organization of the genome and its impact on cellular function, gene expression, and disease progression. Working with Ritambhara Singh and Ghulam Murtaza in the Singh Lab at Brown, he developed scGrapHiC, a deep learning model designed to enhance the accessibility and readability of single-cell genomic data, particularly in the context of chromatin interactions.

“Chromatin, a complex of DNA and proteins, regulates gene expression by altering its 3D structure,” Byron explains. “scGrapHiC uses single-cell gene expression via RNA sequencing (scRNA-seq) to deconvolve complex bulk Hi-C contact maps into refined pseudo-bulk contact maps, making it possible to infer cell-type-specific chromatin interactions without requiring expensive biological assays.”

This approach directly addresses the sparsity and high cost of single-cell Hi-C (scHi-C) data, which often limits insights into rare cell types. The significance of Byron’s work lies in its potential to enable broader, more granular analyses of cell-type-specific chromatin interactions – critical for developmental biology and disease research. Understanding cell-type specific data is particularly important when studying heterogeneous tissues, such as those in tumors and the brain, where distinct cell populations play crucial roles in function and pathology.

Byron’s research has been widely recognized: scGrapHiC was presented at Intelligent Systems for Molecular Biology 2024, where it was nominated for the Ian Lawson Van Toch Best Paper Award, and at Research in Computational Molecular Biology (RECOMB) 2024 as a poster presentation. His work was also published as a full-length paper in Bioinformatics.

Kaleb Newman

Kaleb’s research with Brown CS faculty member Chen Sun tackles a fundamental challenge in AI: recognizing and reasoning about object states – how objects transform physically and functionally over time. While modern Vision-Language Models (VLMs) excel at identifying broad object categories, his findings revealed a crucial limitation: they struggle to distinguish subtle changes, such as recognizing a peeled versus an unpeeled apple.

To address this gap, Kaleb developed ChangeIt-Frames, a dataset of nearly 25,000 images capturing objects in different states, alongside an evaluation framework to rigorously test AI models’ understanding of physical transformations. Through extensive experiments, he found that while models can confidently classify objects, they often fail to recognize their current state, underscoring the need for AI systems that can dynamically interpret the visual world.

“AI models that don’t just see the world but understand it in a dynamic and human-like way will be far more useful and helpful for everyday tasks,” Kaleb says. “This research highlights key knowledge gaps in today’s highly capable models – gaps that must be addressed to create AI that truly perceives the world as humans do.”

Kaleb’s research was accepted to the ECCV 2024 Workshop on Emergent Visual Abilities and Limits of Foundation Models, establishing new benchmarks for AI perception and reasoning.

The full list of Outstanding Undergraduate Researcher Award recipients and honorees is available here.

For more information, click the link that follows to contact Brown CS Communications Manager Jesse C. Polhemus.