Home > Academics > Undergraduate > Undergraduate Resources and Opportunities > Undergrad Research Opportunities > 2023-2024 SURE Research Projects in CSE

2023-2024 SURE Research Projects in CSE

This page lists summer research opportunities in CSE that are available through the SURE Program. To learn more or apply, visit: https://sure.engin.umich.edu/.

Note: CSE does not require additional materials unless noted in the project description.

Directions

Please carefully consider each of the following projects, listed below, before applying to the SURE Program.
You must indicate your top three project choices on your SURE application, in order of preference, using the associated CSE project number.
Questions regarding specific projects can be directed to the listed faculty mentor.
Timeline: SURE applications will be reviewed throughout the month of March and recipients will be notified sometime in late-March/early-April.

Project descriptions

Project #1: Privacy-Preserving Generative AI and Health Analytics
Faculty Mentor: Satish Narayanasamy, [email protected]
Prerequisites: EECS 370; EECS 482 is helpful, but not necessary. Interest in building systems.
Description: Data serves as the foundation upon which generative AI models are constructed. Ensuring privacy for both the data used in training and for the inferences made, as well as protecting the models themselves, is essential to fully unlock the capabilities of generative AI. Numerous important fields like health, finance, and e-commerce require training on private data from various organizations. However, these organizations can’t share data due to privacy worries and regulations (GDPR). For instance, in healthcare, each organization has access to the digital health records of a relatively small group (a few thousands) of patients. If we could combine data from multiple organizations, we could train on millions of patients and create models that accurately predict outcomes for a diverse population. Unfortunately, sharing data among these health institutions isn’t feasible due to clear privacy concerns. We are working on advancing trusted hardware in heterogeneous devices (Intel SGX/TDX, nVidia H100 GPUs, memory, storage), and leverage them to enable privacy-preserving generative AI and health analytics, and ransomware defenses. These confidential computing solutions can guarantee that the data remains in encrypted form, not only during transit and storage, but also during computation when it is being used.
Expected research delivery mode: In-Person

Project #2: Enhancing Observability of Cloud System Software
Faculty Mentor: Ryan Huang, [email protected]
Prerequisites: EECS 482 or equivalent, proficiency in systems programming, EECS 591 (not required)
Description: System software running in cloud infrastructure frequently experience complex gray faults. These complex faults present significant challenges for ensuring the high availability and correctness of cloud systems. This project will develop cross-layer techniques to systematically enhance the observability of modern cloud software. We will explore and design techniques ranging from OS to architecture support, compiler analysis, runtime tracing, and machine learning. Students will work with the faculty and graduate student mentors, and get exposed to latest research as well as hands-on experiences on practical techniques.
Expected research delivery mode: Hybrid

Project #3: Reliable Machine Learning Systems
Faculty mentor: Ryan Huang, [email protected]
Prerequisites: EECS 481 (not required), experience with ML, proficiency in programming
Description: The reliability of ML systems emerges as a crucial concern. Compared to traditional software, ML systems have distinct characteristics that make traditional reliability techniques insufficient. This project, we will investigate real-world bugs and failures in popular ML frameworks and models. Guided by the investigation, we will research advanced testing methods, debugging strategies, and fault-tolerance techniques to comprehensively enhance the resilience of ML systems.
Expected research delivery mode: Hybrid

Project #4: AI for Real-World Emotion Recognition
Faculty mentor: Emily Mower Provost, [email protected]
Prerequisites: An interest in human-centered computing, an interest in machine learning, experience programming in python or experience in another programming language and willingness to learn python.
Description: Emotion, expressed in the real world, is complex and subtle. Yet, when we ask our machines to recognize emotion, we assume that it can be well described by a single word. In this project, you will help to develop new human-centered annotation frameworks for emotion, use crowdsourcing to obtain new labels, and then train large language models (LLMs) to recognize real-world emotion expressions.
Expected research delivery mode: In-Person

Project #5: Speech Emotion Recognition in Noisy Environments
Faculty mentor: Emily Mower Provost, [email protected]
Prerequisites: An interest in human-centered computing, an interest in emotion recognition, an interest in AI, experience programming in python.
Description: Our goal is to understand what happens when we deploy emotion recognition models in real world environments. These models are often developed in controlled laboratory settings. Then, when they are deployed in natural environments, they underperform. We will investigate this issue by measuring how emotion recognition performance changes in noisy environments, how/if these performance changes are also reflected in speech recognition systems, and begin to explore approaches to address these performance changes.
Expected research delivery mode: In-Person

Project #6: Formal Methods for Data Systems
Faculty mentor: Xinyu Wang, [email protected]
Prerequisites: EECS 484 (A or better), EECS 281/280 (A or better), EECS 203 (A or better), very strong programming/engineering skills (e.g., EECS 482 with A or better), self-motivated, interested in research.
Description: Please see the SlabCity paper (on my page: https://web.eecs.umich.edu/~xwangsd/) for more details.
Expected research delivery mode: In-Person

Project #7: Privacy-Preserving Sensing for in-home Activity Recognition
Faculty mentor: Alanson Sample, [email protected]
Prerequisites: Experience with embedded systems, computer vision, or machine learning.
Description: Giving computers the ability to sense and understand our daily activities and routines can enable new smart home applications, context-aware computing, and transform how we detect, monitor, and treat disease and chronic illnesses. However, existing smart devices rely on remote cloud services to process voice commands, analyze video, and perform recognition tasks. Even if these smart devices record only ML features to send to the cloud, private data still leaves the home for cloud-based processing and is stored for an indeterminate amount of time. This project aims to create a new class of smart sensors and embedded devices that removes Personally Identifiable Information before sensitive data leaves the devices while maintaining downstream activity recognition applications. Examples include microphones the remove speech but leave other acoustic information for audio classification and cameras that use onboard GPUs to remove and replace images of people robustly.
Expected research delivery mode: In-Person

Project #8: On-body Ultrasonic Gesture and Touch Interaction Detection
Faculty mentor: Alanson Sample, [email protected]
Prerequisites: Preferred Experience: embedded systems and/or machine learning.
Description: The sensation of touch is one of the fundamental ways we understand and interact with the physical world. However, beyond touchscreens, computing systems have little insight into how users interact with everyday objects. This is particularly important for Augmented Reality systems, which must overlay digital content on the physical world in response to users’ actions. This project aims to investigate novel methods of using ultrasound for on-body sensing of hand gestures, pose, and object interaction events with a focus on creating hardware solutions that are suitable for unobtrusive wearable applications. SURE students will work with a team of undergraduate and graduate students and obtain hands-on experience developing robust embedded systems, real-time programming, and applied machine learning.
Expected research delivery mode: In-Person

Project #9: Robust and Explainable ML for Database System Tuning
Faculty mentor: Lin Ma, [email protected]
Prerequisites: EECS 484 or familiarity with database system internals, experience using database systems (e.g., Postgres, MySQL, Spark) beyond classroom settings (e.g., research or internship projects); experience with machine learning, including neural nets.
Description: Database systems are the central piece of modern data-driven applications. However, they’re notoriously difficult to deploy and optimize due to their complex functionalities and high number of configuration knobs. Recent research and industrial efforts have proposed machine learning (ML)-driven database tuning techniques to automatically optimize the performance of database systems. However, these techniques face the practical challenge that the ML models are often susceptible to out-of-distribution scenarios and their decisions/predictions are hard to reason about. These challenges significantly hinder the broader application of ML-driven database tuning in practice. This project aims to address these challenges by investigating the recent advances in robust and explainable ML techniques and applying them to database tuning problems. This project also seeks to establish a principled evaluation framework that analyzes and establishes the robustness of ML-driven database tuning techniques in production-like environments.
Expected research delivery mode: In-Person

Project #10: Database System Foundation Models
Faculty mentor: Lin Ma, [email protected]
Prerequisites: EECS 484 or familiarity with database system internals, experience using database systems (e.g., Postgres, MySQL, Spark); experience with machine learning, including neural nets (large language models are a plus).
Description: Database systems play an essential role in modern data-driven applications. However, these systems have complex components with a growing list of functionalities that are difficult to configure and optimize. Recent research efforts have proposed machine learning (ML)-driven methods that automatically optimize various components of database systems. Despite the promising results, these methods usually focus on one component in a specific system at a time. Thus, it requires repeated efforts to apply such ML-driven methods from one system to another, including extensive engineering and expensive model training. Inspired by the recent progress from the natural language foundation models (large language models), this project aims to build a database system foundation model that can share its knowledge and intelligence across different database systems. This is possible since database systems often share similar components (e.g., SQL, query planning), though with different implementation tradeoffs. If succeeds, this project can significantly reduce the engineering and training overhead of ML-driven database optimizations, which can have a profound impact. We already made some initial efforts in building the prototype, and we are looking for a new student to help iteratively improve and evaluate our solution, which involves designing and implementing algorithms and running experiments.
Expected research delivery mode: In-Person

Project: #11 Ethical AI Integration in Non-Emergency 911 Services: Shaping Fairness and Reliability
Faculty mentor: Lu Wang, [email protected]
Prerequisites: Fundamental knowledge of natural language processing and AI, proficiency in Python and/or some programming language(s), familiarity with ethical considerations in AI
Description: Our project delves into AI’s rising role in non-emergency contexts for emergency management, notably within 911 dispatch centers, emphasizing ethical guidelines and bias reduction. Centered on these critical services, we prioritize fairness in AI-driven interactions. Leveraging advanced Generative AI and expansive language models, our aim is to begin developing unbiased bots adept at managing non-emergency 911 calls. SURE student(s) will aid in shaping ethical AI integration within these crucial social functions. This project’s contribution will set precedents, ensuring AI operates ethically in vital services. Together, we’re forging a path toward responsible, fair, and reliable AI interactions, transforming the landscape of non-emergency events for the better.
Expected research delivery mode: Hybrid

Project #12: Deploying mobile apps for visually impaired people to access visual information in the real world
Faculty mentor: Anhong Guo, [email protected]
Prerequisites: Skill needed: iOS development (Swift, ReactNative)
Description: Our lab recently developed and deployed two mobile applications for visually impaired people, and they are publicly available on the iOS App Store.
VizLens (https://vizlens.org) is essentially a screen reader that can function in the real world. It reads labels at the direction of the user, who points with their fingers at buttons of interest on control panels. With it, users can employ their smartphone cameras to understand and operate a variety of interfaces in their everyday environments, including home appliances and public kiosks.
ImageExplorer (https://imageexplorer.org), helps visually impaired individuals better understand the content of images. Users can use text-based or touch-based exploration to access images.
Read more about these projects: https://cse.engin.umich.edu/stories/new-apps-for-visually-impaired-users-provide-virtual-labels-for-controls-and-a-way-to-explore-images. In this SURE project, we will improve the functionality of these apps based on user logs and feedback, in order to support more people and needs.
Expected research delivery mode: In-Person

Project #13: OmniScribe: making 360° videos accessible to blind or visually impaired people
Faculty mentor: Anhong Guo, [email protected]
Prerequisites: Skill needed: web development, especially understanding js and APIs.
Description: OmniScribe (https://omniscribe.org/) is a prototype system created to make 360° videos accessible to blind or visually impaired people. OmniScribe enables audio describers to author spatial audio descriptions (AD) for 360-degree videos and allow blind people to consume them through a mobile prototype. We have proved OmniScribe can help audio describers to better author ADs for 360-degree and help blind people engage in the video immersively. However, what we conducted were lab-controlled studies, which might create biases. To understand blind users’ real feedback on the spatial ADs, in this SURE project, we will create a web 360-degree player and mechanism to collect their responses, where blind people can access the descriptions OmniScribe enabled.
Expected research delivery mode: In-Person

Project #14: Hardware-Software co-design for Nuclear SecurityFaculty mentor:
Faculty mentor: Reetuparna Das, [email protected]
Prerequisites: EECS 370, EECS 281
Description: This project will explore state of art domain specific hardware acceleration techniques for HPC algorithms in the domain of nuclear security.
Expected research delivery mode: In-Person

Project #15: Advancing Internet shutdown resistance technologies
Faculty mentor: Roya Ensafi, [email protected]
Prerequisites: EECS 388 and EECS 482
Description: Traditional censorship primarily involves blocking access to specific websites or online services. However, a concerning trend has emerged with the escalation of a more extreme form of censorship—deliberate Internet shutdowns, resulting in complete disconnection and significantly impacting lives in affected regions. Addressing this issue requires exploring both non-academic and limited academic solutions designed to facilitate access during such shutdown events. The overarching goal of this project is to compile and analyze these tools, categorizing them based on their capabilities, threat models, availability, and usability. Additionally, we aim to identify and address gaps in implementing shutdown resistance technology across various stakeholders, including end-users, tool developers, and non-governmental organizations. To achieve this, the project will leverage extensive user studies and interviews. Ultimately, the outcome will be a practical, shutdown-resistant solution accompanied by guidelines for the development of similar tools in the future. The student involved will collaborate closely with the professor and other lab students to successfully navigate each stage of the project.
Expected research delivery mode: In-Person

Project #16: Leveraging Reinforcement Learning to Automate the Generation of Filter Rules for Adblocking
Faculty mentor: Roya Ensafi, [email protected]
Prerequisites: EECS 388
Description: The advertising and tracking industry is an integral part of our digital lives, and yet its aspects often compromise user privacy. Millions of users worldwide rely on privacy-enhancing tools, such as adblockers, to block and hide content associated with ads and tracking. Yet the efficacy of these tools rely on manually-maintained and error-prone filter rules. Our project aims at revolutionizing this landscape. The project embarks on the usage of reinforcement learning techniques for generating and evaluating filter rules. The goal is to efficiently block ads and tracking, while minimizing visual and functionality breakage on websites. As our project develops, we envision a future of automatically curated filter rules that can protect user privacy worldwide across different platforms beyond the web. SURE students working on this project will design and implement different algorithms for filter rule generation, as well as learn about and integrate state-of-the-art prior works into our framework. They will have opportunities to extend the reinforcement learning method, scale and automate the entire system using tools like Selenium, Docker, and cloud platforms.
Expected research delivery mode: In-Person

Project #17: Censored Planet Censorship Observatory
Faculty mentor: Roya Ensafi, [email protected]
Prerequisites: EECS 388
Description: The Censored Planet observatory uses a modular design for measuring and analyzing Internet censorship. Censored Planet continuously measures reachability to 2,000 websites from more than 95,000 vantage points in 221 countries. The observatory was launched in August 2018, and has since then collected more than 45 billion measurement data points. Censored Planet uses features in existing Internet protocols and infrastructure to interact with remote systems, using their responses to determine the presence or absence of censorship. Building off of novel Internet measurement and Machine Learning techniques, SURE students will work with faculty and graduate student mentors to implement new features into the Censored Planet observatory, and ensure sustainability of current features.
Expected research delivery mode: In-Person

Project #18: VPNalyzer
Faculty mentor: Roya Ensafi, [email protected]
Prerequisites: EECS 388
Description: VPNalyzer is an interdisciplinary research project from the University of Michigan that aims to analyze the VPN ecosystem. VPNalyzer consists of three parallel efforts: large-scale quantitative and qualitative user studies, a cross-platform desktop tool for users to test the security and privacy features of their VPN connection, and qualitative studies surveying VPN providers. Our goal with the VPNalyzer project is to advance the public interest, inform practical regulations and standards, enforce accountability and empower consumers to find more trustworthy VPN products. Expanding the existing large-scale study of VPN providers, SURE students will work on analyzing data collected from the beta release of the VPNalyzer tool from users around the world and work on updating the VPNalyzer website.
Expected research delivery mode: In-Person

Project #19: Quantum computing cloud management stack
Faculty mentor: Ang Chen, [email protected]
Prerequisites: Experience building large software systems (10K+ LoC in systems languages like C/C++/Go)
Description: To maximize quantum device utility and quantum application fidelity, we seek to enable quantum computers for all by developing the gatekeeper of the quantum cloud: a quantum cloud resource manager. We call this vision Qube, analogous to the widely-used classical resource manager Kubernetes, or Kube.
Expected research delivery mode: In-Person

Project #20: Program lifting for cloud infrastructure-as-code
Faculty mentor: Ang Chen, [email protected]
Prerequisites: Formal methods. Systems prototyping. Cloud familiarity.
Description: Program verification and synthesis using LLMs. The target is cloud infrastrucure-as-code (IaC) such as Terraform code. We want to generate IaC programs from execution traces in cloud providers.
Expected research delivery mode: In-Person

Project #21: Classical computing research to advance quantum computing
Faculty mentor: Gokul Ravi, [email protected]
Prerequisites: Some background in computer architecture, quantum computing, Python programming is preferred but not required.
Description: Quantum computing is a disruptive technological paradigm with the potential to revolutionize computing and, therefore, the world. Therefore, as quantum devices transform from laboratory curiosity to technical reality, we must unlock the full potential of quantum computing to achieve meaningful benefits on real-world applications with imperfect quantum technology. The opportunities for classical computing based research to support quantum computing are vast: technology-aware circuit compilation, classical pre/post-processing for quantum executions, latency/bandwidth/power constrained design, scalable classical simulation, multi-chip strategies, cloud resource management, feedback-based optimization, benchmarking and design space exploration, and much more. Students will work with the mentor to identify research topics in one of these areas that fit well with their background and interests.
Expected research delivery mode: In-Person

Project #22: Pulse computing
Faculty mentor: George Tzimpragos, [email protected]
Prerequisites: Prior experience in programming, digital logic, and computer architecture design is required. Students with a high level of curiosity in a breadth of subjects (for example, students pursuing double majors or minors, or those that generally have interdisciplinary interests) will be preferred.
Project Description: Digital logic design, as we know it since its inception, comes with the requirement that latching switching elements are used. In some cases, however, the future of computing may rely on devices–from optical and superconducting to biological–that, unlike relays, tubes, and transistors, cannot remain in an On or Off state for arbitrary amounts of time. This project aims to investigate computing with transient pulses by looking at ways to encapsulate pulses’ interaction mathematically and define systems for their effective manipulation. The students involved will 1) review relevant literature, 2) analyze the emerging trade-off space, and 3) devise microarchitectures built on recent logic and memory advancements.
Expected research delivery mode: In-Person

Project #23: Full-stack quantum system simulation infrastructure
Faculty mentor: George Tzimpragos, [email protected]
Prerequisites: Prior experience in programming and digital logic is required. Additional experience in quantum computing and programming languages would be desired.
Project Description: Even with recent advancements in superconducting qubit fabrication, realizing a practical error-corrected quantum computer demands over one million physical qubits. However, supporting thousands or millions of qubits solely with traditional CMOS quantum-classical interface technology is not feasible due to thermal and latency limitations. Consequently, future quantum systems will need to embrace a more heterogeneous architecture, incorporating both CMOS (and cryo-CMOS) and digital superconducting elements into their design. Building upon our group’s prior work on domain-specific languages for hardware design, this project aims to develop a cross-technology simulation framework for full-stack quantum system microarchitectures.
Expected research delivery mode: In-Person

Project #24: Verification-first hardware design
Faculty mentors: George Tzimpragos & Yatin Manerkar, [email protected] & [email protected]
Prerequisites: Prior experience in programming and digital logic is required. Additional experience in formal methods would be a plus.
Project Description: Computer systems have become ubiquitous in every aspect of our lives. However, their vulnerabilities expose users to potential attacks, information leakage, and even risks to physical safety. In response to these concerns, formal methods in computer architecture have emerged as essential tools for reasoning about microarchitecture and hardware implementation, ensuring system correctness and security. This project seeks to address scalability limitations in current formal verification approaches by unifying linear temporal and race logic, bridging the semantic gap between computation and verification operators. The ultimate objective is to foster a paradigm shift towards “verification-first” hardware system design.
Expected research delivery mode: In-Person

CSE Project #25: Query Optimization
Faculty mentor: Xinyu Wang, [email protected]
Prerequisites: EECS 484 (or familiarity with SQL), or EECS 481 (software engineering), or EECS 483 (compilers).
Description: SQL queries, if written poorly, are slow on large databases, even using state-of-the-art query optimizers. This project aims to develop a super optimizer for SQL queries, which is able to maximally boost the performance of a poorly written query.
Expected research delivery mode: Hybrid but likely with more in person meetings

CSE Project #26: Usable Sound Awareness Systems for Deaf and Hard of Hearing People
Faculty mentor: Dhruv Jain, [email protected]
Prerequisites: Some experience with implementing machine learning (ML) algorithms is preferred. Experience with designing front-end user interfaces and/or conducting user studies is a huge plus, and if you have strong experience with this, but are uncomfortable with ML, please still apply.
Description: We will research, build, and deploy systems to provide sound awareness to people who are deaf and hard of hearing. These could include, but not limited to: (1) augmented-reality speech transcription interfaces on head-mounted displays, (2) smartwatch-based sound recognition app, and (3) a web-based personalizable sound recognition system. You will be part of a next-generation team who is actively working with Deaf/disabled population with a history of successful product launches. Once your system is built, you will participate in conducting field studies with DHH users and help with open-sourcing your system, ultimately leading to a huge real-world impact.
Expected research delivery mode: In-Person

CSE Project #27: Computational Strategic Reasoning
Faculty Mentor: Michael Wellman, [email protected]
Prerequisites: Programming ability; interest/background in finance, economics, game theory, and/or statistics (helpful though not required)
Description: The Strategic Reasoning Group (strategicreasoning.org) develops computational tools to support reasoning about complex strategic environments. Recent applications include scenarios arising in finance and cyber-security. We employ techniques from agent-based modeling, game theory, and machine learning.
Expected research delivery mode: In-person

CSE Project #28: Hazel: A Live Functional Programming Environment
Faculty Mentor: Cyrus Omar, [email protected]
Prerequisites: EECS 490 or equivalent is preferred, but not required.
Description: Hazel (hazel.org) is a live functional programming environment that is able to typecheck, transform and even execute incomplete programs, i.e. programs with holes. There are a number of projects available within the Hazel project for a student interested in research into programming languages, both theoretical and human-centered in nature.
Expected research delivery mode: In-person preferred, virtual possible

CSE Project #29: RustViz: Interactively Visualizing Ownership and Borrowing
Faculty Mentor: Cyrus Omar, [email protected]
Prerequisites: EECS 490 or EECS 483 or equivalent is preferred, but not required.
Description: Rust is unique in that it is a memory-safe and thread-safe programming language that does not use a run-time garbage collector. Instead, it enforces a static ownership and borrowing discipline (“borrow checking”) to ensure that resources can be managed fully statically. However, there is a learning curve when programmers first encounter Rust’s new ideas. This project will contribute to the RustViz project, which is developing a visualization system for Rust’s ownership and borrowing semantics. Possible projects include a system for deriving this visualization directly from Rust compiler internals, designing a new more scalable visual language, or visualizing more advanced features like region-based memory management.
Expected research delivery mode: In-person preferred, virtual possible