Fung Institute End of Year Showcase 2025

Team: Charitha Kaki [MSE], Charles Liu [MSE], Tyler Zierer [ME]
Advisor(s): Hayden Taylor
Project ID: 123

Metal Organic Frameworks (MOFs) are a class of materials which are made up of a metallic component and an organic linker component. This structure allows the molecules to be extremely tunable for different applications. One particular copper-based MOF, named HKUST-1, is tuned specifically to be able to absorb carbon dioxide due to the size of the pores created by the chemical framework. HKUST-1 can then be “regenerated” by removing the trapped carbon dioxide through a variety of processes, typically using a pressure or temperature swing. Our team hopes to regenerate MOFs using mechanical deformation, or more specifically, axial strain. Our hypothesis is that by stretching the MOF, the pore size may change enough to break the affinity between the carbon dioxide and MOF structure without plastic deformation. If successful, this would prove to be a far more energy efficient method of regenerating MOFs and capturing atmospheric carbon dioxide.

Team: Kaviyasri Saravanan [ME], Vineeth Parashivamurthy [ME], John Schofield [ME], Anna Sajer [MSE], Sebastian Marini [MSE]
Advisor(s): Peter Hosemann, Alexandros Spyromilios
Project ID: 148

Diamonds have excellent thermal and mechanical properties that can be leveraged for advancements in space, medical, electronic, and quantum computing applications. By laser texturing diamonds, we can engineer the synthetic diamond surface to feature periodic nanostructures and potentially uncover distinct surface properties. Our team is focused on understanding and optimizing the specific process conditions in which these structures form. Using various materials characterization techniques, we aim to identify optimal experimental parameters to replicate these nanostructures for advanced materials performance.

Team: Emilia Garcia-Bompadre [BIOE], Ganchai Siriwatcharapibool [BIOE], Joseph Lee [BIOE], Ragini Vidyashankar [BIOE]
Advisor(s): Chris Anderson
Project ID: 49

Bioproduction has immense potential to revolutionize small molecule production by providing a sustainable alternative to chemical synthesis, but its scalability is limited due to high costs and process inefficiencies. This is due to microbial cells prioritizing resources for growth over product formation, generating up to 300 million liters of biomass waste annually. Our team is engineering E. coli to develop environmentally controlled digestion enzymes that reformulate spent microbial biomass into nutrients to feed future cell generations. This approach reduces waste and enhances productivity, improving yield and economic viability of bioproduction.

Team: Randi Roy [BioE], Abhay Bhat [BioE], Trisha Andrews [BioE], Roshan Sen [BioE]
Advisor(s): Chris Anderson
Project ID: 52

Atom-mapping is like a GPS tracker for chemical reactions, as it tracks the correspondence of atoms in the reactant to the product. Current atom-mapping methods deployed in chemoinformatics utilize mathematical frameworks that lack the underlying chemical logic, resulting in low accuracy. MoeityMapper takes a unique approach incorporating this chemical logic by first simplifying complex molecules to their carbon backbones, using this as a basis to map each atom more accurately than current approaches. This tool is crucial in pharmaceutical development and chemical research, enabling researchers to understand molecular transformations and improve drug design and safety.

Team: Elisabeth Young [MSE], Hunter Joerger [BioE], Zubin Havewala [BioE]
Advisor(s): Seung-Wuk Lee, Fiona Doyle
Project ID: 93

~90% of the rare earth elements (REEs) used in the US are imported. While they are indispensable in consumer electronics, electric vehicles, and renewable energy, the EPA restricts hazardous waste from conventional REE refining, limiting domestic production. Our team is developing a membrane fabricated from fusion proteins that combine Hans-Lanmodulin, which binds REEs, with elastin-like polypeptide. The protein is purified from bacteria, drop-casted, and crosslinked to form a pH-responsive membrane that facilitates selective transport of REEs. Eco-friendly purification technologies have the potential to enable a secure, environmentally responsible domestic supply of critical materials that support technological innovation and national security.

Team: Qinyi Xu[ME], Brooke Smith[ME], Deena Kumar[ME], Ropafadzo Nhanga[ME], Edmond Rozan Loubeyre[ME]
Advisor(s): Peter Hosemann
Project ID: 125

Brake discs made of cast iron are easier to machine and process when the material ages for approximately one week after casting. Aged cast iron improves tool life during machining resulting in better surface quality of the brake discs. In collaboration with Brembo, we want to understand the metallurgical changes that happen during the aging process. The aging process will take place under three temperatures over time to determine when it occurs. Characterization tests on fresh cast iron will identify chemical and microstructure changes. The goal is to reduce the aging time under a week for faster brake manufacturing.

Team: Juyi Zhang [MSE], Langjian Zhu [BIOE], Maximillian Dean [BIOE], Emily Arana Barcala [BIOE]
Advisor(s): Iain Clark
Project ID: 128

Our project tackles the organ shortage by developing a 3D bioprinting system that utilizes electric fields to precisely position living cells, eliminating the need for moving parts commonly seen in traditional bio-active 3d printers. We engineered components to enhance stability and functionality and created software that is capable of translating design files into precise printing instructions. This technology advances the creation of lifelike tissues and brings us closer to patient-specific bioprinted organs.

Team: Ayub Khan [ME], Mohamed Cherif [MSE], Nathan Lee [ME], Patrick Bayo [ME], Yuran Wen Wen [MSE]
Advisor(s): Fei Teng
Project ID: 133

We propose an innovative air-free container design featuring vacuum pumping or protective gas filling via top holes, a threaded base for airtightness, and a central pillar to secure samples or oxygen absorbers. Made of PMMA with rubber seals, the design ensures effective sealing at low cost.

Team: Yiran Yuan [ME], Yuchao Huang [ME], Zike Yan [MSE], Ji Su Bok [ME]
Advisor(s): Tarek Zohdi
Project ID: 155

Multi-layer insulation (MLI) blankets are used as a passive thermal control to protect satellite components. With the increasing number of satellites being launched, MLI faces challenges in large-scale production due to labor-intensive manual sewing and brittle failure from tearing and creasing. This project investigates an alternative approach to enhance thermal control and align with industrialized processes used on satellite assembly lines. Preliminary finding
suggests integration of blanket design algorithms with innovated materials reduces the manufacturing time and structural failure.

Team: Tristan Bourgade [ME], Connor Vidmar [ME], Evan Percival [ME], Scarlett Hao [ME], Erfan Kohyarnejadfard [ME]
Advisor(s): Taylor Waddell, Hayden Taylor
Project ID: 156

This project aims to make computed axial lithography (CAL), a new 3D printing technology, more accessible and affordable. Developed at UC Berkeley and a few other laboratories, CAL has not been standardized, and each lab has its own version. Our goal is to redesign the system to be easy to use and cost less than half of the current price to produce. We will also address safety and usability issues, making it easier for more people to get involved in CAL research. By sharing our work with the open-source community, we hope to accelerate the development of this promising technology.

Team: Maggie Gao [IEOR], Jonathan Yu [BioE], Stephan Alfaro [BioE]
Advisor(s): Chris Anderson
Project ID: 48

The process of producing pharmaceuticals in traditional settings requires significant capital and resources. Currently, biopharmaceutical manufacturing is mostly concentrated in highly developed countries, leaving poorer nations dependent on importing medications, often with limited access to essential resources. Our team has developed a consumer-friendly Escherichia coli bioreactor designed for ease of maintenance, distribution, and manufacturing, all while being extremely affordable to operate. This innovative bioreactor system prioritizes a scale-out approach rather than a scale-up method, minimizing the initial resources and costs typically required for biopharmaceutical production.

Team: Sandya Wijaya [IEOR], Jacob Bolano [EECS], Shriyanshu Kode [EECS], Yue Huang [EECS], Alejandro Gomez Soteres [ME]
Advisor(s): Anant Sahai
Project ID: 104

Large Language Models (LLMs) often struggle with code generation tasks involving niche software libraries. Existing code generation techniques with only human-oriented documentation can fail — even when the LLM has access to web search and the library is documented online. To address this challenge, we propose ReadMe.LLM, LLM-oriented documentation for software libraries. By attaching the contents of ReadMe.LLM to a query, performance consistently improves to near-perfect accuracy, with one case study demonstrating an improvement form 20% to 100% success across all tested models. We propose a software development lifecycle where LLM-specific documentation is maintained alongside traditional software updates. In this study, we present two practical applications of the ReadMe.LLM idea with diverse software libraries, highlighting that our proposed approach could generalize across programming domains.

Team: Xiang Feng [IEOR], Chen-Ching Lin [ME], Kate Barouch [BIOE], Shaojie Wang [ME], Willem LARRAS [ME]
Advisor(s): Gabriel Gomes
Project ID: 36

Our team explored machine learning applications across various engineering disciplines, focusing on classification to solve engineering challenges. Through peer-driven learning and troubleshooting, we enhanced our individual project outcomes while refining our technical and communication skills. By sharing insights and learning from one another, we deepened our understanding of machine learning and its practical applications. Our goal is to bring this experience into our future engineering roles, contributing to innovation and problem-solving in society.

Team: Yuan-Hao Huang [EECS], James Peng [EECS], Kevin Shen [ME]
Advisor(s): Murat Arcak, Hanna Krasowski
Project ID: 72

Every year, hundreds of ships collide at sea. More than half of these tragedies are due to human error and should be easily avoidable. We introduce the AI Captain: an AI model that observes a seaborne vessel’s surroundings and adjusts its trajectory if it’s about to collide with an obstacle. We have developed an algorithm to find scenarios with ships that have close encounters from real-world data and have trained various reinforcement learning models to navigate past these obstacles without collision. We use CommonOcean, a leading framework for motion planning of autonomous vessels, to build an end-to-end pipeline for learning.

Team:Sumeet Kulkarni [EECS], Luca Petrescu [ME], Rishi Kumar Srinivasan[CEE], Junhua Ma [EECS]
Advisor(s):Murat Arcak
Project ID: 64

Classical controllers for Vertical Takeoff and Landing (VTOL) aircraft lack adaptability and struggle to handle the complex, nonlinear dynamics of these vehicles. While Reinforcement Learning models offer great promise in addressing the limitations of classical control methods, there has been resistance from the industry and regulatory agencies due to the its unpredictability. Hence, the implementation of stability margins bring RL control closer to application. We are developing a neural network controller that combines the guaranteed stability margins of classical control with the adaptability of neural networks. To achieve this, we employ a novel reinforcement learning algorithm that ensures the model satisfies the desired safety metrics during training.

Team: Ali Shazal [EECS], Samuel Sovi [EECS], Zhongxing (Ryan) Zhou [IEOR], Jasper Liu [EECS]
Advisor(s): Reehan Shah (Asurion), Gabriel Gomes (ME)
Project ID: 153

Third-party hosted Large Language Models (LLMs) in customer support face challenges such as slow response times, data privacy concerns, and frequent downtime. To overcome these limitations, we are collaborating with Asurion to replace their reliance on OpenAI’s GPT-4o with fine tuned, open-source LLMs. By applying prompting and tuning techniques, our optimal setup achieves 96% accuracy with an average latency of 0.5 seconds, making it 3 times faster than GPT-4o without sacrificing result quality.

Team:Dinesh Rajasekaran [IEOR], Jing Tang [CEE], Wenbin Wan [IEOR]
Advisor(s): Ivan Bondarenko
Project ID: 91

Our project, Pandora’s Box, is a smart logistics container designed to enhance supply chain transparency and efficiency. Equipped with IoT sensors, GPS tracking, and RFID technology, it enables real-time monitoring of location, condition, and contents. The container seamlessly integrates with existing logistics systems, including the Woosh transportation network, and is designed to fit multiple vehicle types for maximum flexibility. By improving asset visibility and security, Pandora’s Box helps reduce delays, prevent losses, and optimize operations across the supply chain.

Team:Chang Su [IEOR], Siqing [IEOR], Xiaoya Zhu [IEOR], Chunxiao Zhang [IEOR]
Advisor(s): Paul Grigas
Project ID: 95

Creating scalable, data-driven solutions that improve operational efficiency, costeffectiveness, and customer satisfaction on real-world datasets from JD.com. We leveraged machine learning, deep learning and optimization techiniques to investigate user behavior, revenue trend, warehouse allocation and delivery routing problems.

Team: Akshay Trikha [MSE], Kyle Chu [EECS], Advait Gosai [EECS], Parker Szachta [EECS]
Advisor(s):Matthew Sherburne
Project ID: 109

Predicting material properties is crucial for designing better batteries, semiconductors, and medical devices. Deep learning helps scientists quickly find promising materials by predicting their energy, forces, and stresses. Companies scale capacities of deep learning models in multiple domains, such as language modeling, and invest many millions of dollars into such models. Our team analyzes how scaling training data (giving models more information to learn from), model sizes (giving models more capacity to learn patterns), and compute (giving models more computational resources) for neural networks affects their performance for materials property prediction.

Team: Yiyun Zhang [IEOR], Kexin Tang [IEOR], Shreejal Luitel [IEOR]
Advisor(s): Samuel Gonzalez
Project ID: 22

Today, financial firms leverage companies’ networks of suppliers, partners, competitors, and investors to identify potential risks. This project focuses on developing an optimal procedure for extracting supply chain relationships using large language models (LLM) by reading unstructured data sources for entity linkage and network analysis. This enables firms to proactively manage risks in their ecosystems and informs investment decisions.

Team: Shujie Deng [EECS], Xinran Fang [EECS], Lisa (Qi) Hou [EECS]
Advisor(s): Allen Y. Yang
Project ID: 7

Imagine controlling a robot on Mars by just talking to your phone. Our team is developing an Android app that enables users to command and monitor robots directly using natural language, eliminating the need for complex manual controls. By integrating a locally deployed large language model, real-time telemetry visualization, and voice recognition, our solution showcases the power of AI-driven human-computer interaction and the capabilities of mobile AI chips in real-world applications.

Team: Ram Krishnamoorthy [IEOR], Jeremy Fischer [EECS], Vishal Kumar [EECS]
Advisor(s): Mahdi Al-Husseini
Project ID: 149

The US Army is looking to create a sense of realism in decision-making under pressure while training their Medical Evacuation (MEDEVAC) personnel to plan for mass casualties. We are creating a 3-D Medical Evacuation Wargame over multiple environments and scenarios to allow MEDEVAC students to control various platforms like Blackhawks, Medical ships, and Ground vehicles for real-time treatment coordination. Our solution uses the game engine Unity, net code for game objects, ML agents to create an adversarial casualty generator, and intense military knowledge to develop rigorous Operation Orders and Road2Wars to emphasize realism.

Team: Daniel He, Rachel Zeng, James Hu
Advisor(s): Sanjit Seshia
Project ID: 116

Traditional physical training methods are costly and inaccessible, averaging $50-$100 per session and requiring travel, while alternatives such as online tutorials lack personalization and exercise monitoring for effective and safe at-home practice. To address these challenges, our team is developing an innovative mixed reality (MR) system that provides personalized, interactive training experiences. By integrating MR devices with AI-driven real-time dynamic scenario generation, our system enables trainers without technical knowledge to create training tutorials personalized for individual trainees. Our solution simulates in-person training, making physical instruction more affordable, accessible, and engaging.

Team: Sophie Pineau [NE], Naomi Ovrutsky [ME], Ace Meng [NE], Jacob Sitemo [ME]
Advisor(s): Lee Bernstein
Project ID: 43

Accurate measurements of how 14 MeV neutrons interact with nuclei are crucial for fields like nuclear security, space exploration, and national security. Currently, these measurements have large uncertainties due to challenges in accurately determining neutron flux. We helped assemble a new DT-API (Deuterium-Tritium Associated Particle Imaging) neutron source on the UC Berkeley campus, which aims to minimize this uncertainty by precisely determining neutron location. We also participated in a measurement campaign led by researchers from Johns Hopkins Applied Physics Laboratory, Lawrence Berkeley National Lab (LBNL) and NASA to measure these data using an existing DT-API system.

Team: Shobhit Brijesh [ME], Jay Darji [MSE], Matthieu Dagousset [NE]
Advisor(s): Max Monange
Project ID: 65

By 2050, global energy demand is projected to double, while existing energy sources remain unsustainable. Nuclear fusion promises a path to unlimited, clean, and scalable power, paving the way for a future free from resource constraints. Our team is pioneering a comprehensive damage modeling process to evaluate candidate materials for the plasma-facing wall of an inertial fusion reactor. By leveraging advanced simulation tools, we analyze fluid flow, thermomechanical stress, and irradiation resistance to optimize first wall design. Our objective is to provide EX-Fusion Inc. with robust recommendations for future reactor applications, ultimately advancing nuclear fusion designs to power a world with abundant energy.

Team:Noor AL-Sulaiman [ME], Paul Jin [Me], Yichen Hu [ME]
Advisor(s):Michael Gollner
Project ID: 41

Over 57,000 acres of land in total per year in California are lost to wildfires, understanding plant flammability is essential for wildfire prevention. Calorimeters provide critical fire-risk data, but existing models are expensive and inaccessible. The combustion crew team is developing a low-cost, scalable calorimeter to analyze gas composition and assess flammability with high accuracy. Our design integrates real-time weight measurement, optimized exhaust flow, and refined gas sampling. By leveraging load cells, an advanced exhaust system, and simulation tools, we ensure precise combustion monitoring. This work enhances accessibility to fire-risk analysis, providing an affordable tool for researchers, educators, and analysts.

Team:Kai Schuster [MSE], Meghana Mahendra [ME], Victor Malbrel [ME]
Advisor(s):Vianney Grenez
Project ID: 20

AI has created an insatiable demand for data center energy. The biggest contribution to this demand is the cooling that is required to cool the chips down. Traditional approaches, such as air cooling, are unable to provide the thermal performance for today’s workloads, and they are energetically inefficient. Liquid cooling technologies are being implemented in order to meet the demand. One of those technologies is immersion cooling. In immersion cooling, the electronics are fully submerged in a dielectric fluid to cool them. These dielectric fluids are expensive, toxic, and have poor thermal performance. In this project, we aim to explore the substitution of water for dielectric fluids in immersion cooling. In order to accomplish this, we coat the electronics in a thin, dielectric coating to electrically protect them from shorting to the water, and submerge them in an custom, automated test setup that we designed and built.

Team: Reed Harris [CEE], Shuhang Zhong [MSE], Jiahao Yao [MSE]
Advisor(s): Ana Arias
Project ID: 21

Current agricultural practices lack data for optimizing fertilization and watering, leading to nitrate runoff and environmental contamination. TerraSync tackles these problems with an innovative electrochemical soil sensor for continuous monitoring of nitrate concentrations over a growing season. As part of this system, we have developed a cheap printable Ag/AgCl reference electrode with enhanced stability and extended lifetime—designed to be both durable and robust. While developed for agriculture, this technology holds value across any field requiring a low-cost reliable reference electrode.

Team:Esteban Labrador [ME], Sean Shitamoto [ME], Sascha Turovskiy [NE]
Advisor(s): Guanyu Su, Ben Li
Project ID: 42

Tritium is a fuel source for nuclear fusion devices. Manufacturing tritium is highly expensive as it is produced typically in nuclear reactors, so modern fusion energy devices utilize molten salt to produce tritium. The issue comes from extracting that tritium out of the salt such that it can be used for future fuel cycles. Our team is using computational fluid dynamics software to analyze various heat exchanger designs to determine a cost-effective method of tritium extraction from such molten salt.

Team: Rianna Campbell [BIOE], Samuel Xu [BIOE], Dominique de Fiesta [BIOE]
Advisor(s): Syed Hossainy
Project ID: 136

Our team is developing a robust computational framework to optimize the geometry of drug-delivery particles injected into the bloodstream in order to enhance their localization to aneurysms. By simulating physiologically accurate conditions and applying machine learning, we aim to advance safer, minimally invasive treatments.

Team: Abdul-Jabbar Mohammed [BIOE], Joseph Accurso [BIOE], Lauren Hunter [BIOE], Mark Provost [BIOE], Raj Thimmareddy [ME]
Advisor(s): Boris Rubinsky
Project ID: 10

MEGAN (Machine-learning Enhanced Gesture-based Analysis of Neurodegradation) is a wearable system designed to quantify early cognitive and motor deficiencies associated with Parkinson’s disease. By integrating IMU, sEMG, and Computer Vision data during a structured gesture exam, MEGAN enables a multimodal machine learning framework to detect subtle impairments before symptoms become clinically apparent. The project aims to shift Parkinson’s care toward proactive, data-driven intervention.

Team: Douglas Hutchings [ME], Ginghei Mao [ME], Aakanksh Bhat [ME]
Advisor(s): Tiffany Tao, Gabriel Gomes
Project ID: 45

190,000 stroke survivors annually in the United States will suffer long-term mobility impairment and require rehabilitation.
Our team, in partnership with Blue Goji, applies modern robotics control techniques to medical treadmills, thereby promoting access to physical rehabilitation.
By integrating cameras, force sensors, and accelerometers, our intelligent treadmill safely adapts to the user, supplementing the efforts of physical therapists.

Team: Peter Nguyen [BioE], Alice Zhang [IEOR], Lucy Liu [IEOR], Neehar Thumaty [BioE], Sebastian Mejia [EECS]
Advisor(s): Tiffany Tao, Wael, Elise Scipioni, Coleman
Project ID: 47

Our team explores whether Blue Goji’s gamified exercise bikes can enhance college students’ mental well-being and stress relief through interactive workouts. By merging exercise and gaming, we assess improvements in motivation, emotional resilience, and overall mental health.

Team:Aryan Pammar (BIOE), Jordan Snyder (BIOE), Phung Le (BIOE)
Advisor(s): Shawn Shadden
Project ID: 53

Through the use of deep learning, automatic segmentation of blood vessels from CT and MRI scans has the potential to allow cardiologists to avoid the tedious process of manual identification and devote more time to analysis. However, it currently cannot replace manual segmentation in a clinical setting due to a lack of high quality training data. We introduce a scalable framework for improving vessel segmentation in CT scans by addressing the challenges of manual annotation and limited labeled data. We optimize SeqSeg, a deep learning tool built on nnU-Net, using a closed-loop training process where validated model predictions are added back into the training set—enabling automatic dataset expansion without additional manual labeling. The model achieves high accuracy and improved generalizability, with promising potential for clinical use.

Team: Aidan Dunne, Xiangyu Sha, Zsofia Keszei
Advisor(s): Kadidia Konate
Project ID: 75

Hair stylists often work for 8-10 hours braiding the hair of a single customer, which causes repetitive stress injuries like carpal tunnel, trigger finger, and back pain. Our team’s goal is to design an automated hair-braiding device that revolutionizes the industry by significantly reducing braiding time, improving stylist working conditions, and enhancing salon efficiency. Using an innovative diving bell-inspired hair-holding mechanism and a maypole braider, we aim to automate the repetitive part of hair braiding.

Team: Cyrill Castr [BIOE], Jonathan Lyda [BIOE], Justin Sung [BIOE], Gracie Teed [BIOE]
Advisor(s): Angela Rizk-Jackson, Gundolf Schenk, Yongmei Shi & Gabriel Gomes
Project ID: 76

Genetic researchers and clinicians must enhance their understanding of the human genome to improve diagnosis and develop targeted treatments, but existing variant prediction tools can be inaccessible. PathoGenX bridges the technology gap between users and Meta’s ESM1b, an open-sourced transformer protein language model, enabling more accessibility to accurate genetic variant analysis. With automated data processing, improved pathogenicity prediction, and annotation, our solution simplifies genomic analysis and expands access to powerful variant interpretation tools, maximizing breakthrough opportunities.

Team: Adam Johnson [BIOE], Jeremie Wisniewski [BIOE], Victoria Yang [BIOE]
Advisor(s): David Krucik, Xina Quan, Art Muir, Gabriel Gomes
Project ID: 78

Veterinary medicine lacks affordable, accurate, and non-invasive blood pressure monitoring tools suited to clinical needs. To address this, we worked with medical device company PyrAmes to adapt their continuous, non-invasive blood pressure (cNIBP) technology from the human to the veterinary market. Through market analysis and interviews, we identified reusability as the top design priority. We proposed two key engineering changes: (1) a redesigned surface connection between the sensor and electronics, and (2) a rechargeable battery system. Our solution enables accurate, continuous, and cost-effective blood pressure monitoring for veterinarians under budget limitations.

Team:Kamand Shafieha [BIOE], Anushka Joshi [BIOE], Sruthi Ponnambalam [BIOE], Jason Bie [BIOE]
Advisor(s): Lee Fleming
Project ID: 9

The current cancer treatments in the market have the potential to be more targeted and cause fewer adverse effects for the patients. Lawrence Berkeley National Laboratory is currently working on an efficient production pipeline for Actinium-225, a promising medical isotope for cancer treatment. Utilizing our bioengineering background and our understanding of the pharmaceutical industry, our team aims to facilitate this commercialization process by establishing relationships with the business stakeholders, managing risk, formulating business models, and providing financial insights to the technical team.

Team:Aalaya Wudaru [BIOE], Jash Gujarathi [EECS], Ryan Ha [BIOE], Yiying Lu [EECS], Smrithi Surender [BIOE]
Advisor(s): Liwei Lin
Project ID: 30

Nearly half of American adults have hypertension, yet most remain unaware, increasing their risk for heart disease and stroke. Current monitoring methods either lack accuracy and continuity (like standard cuffs), or require invasive procedures unsuitable for routine use. This underscores the need for a non-invasive, continuous blood pressure monitoring solution. Our device combines optical and ultrasonic sensors with personalized calibration to generate real-time blood pressure waveforms. By offering high-fidelity, continuous insights, our novel device empowers earlier detection and more effective management of cardiovascular health.

Team:Allegra Messinese [BioE], Jacqueline Mejia [BioE], Yizhen Jia [BioE], Ali Habbal [BioE]
Advisor(s): Syed Hossainy, Dorian Liepmann, Michael Conboy
Project ID: 11

Cardiovascular diseases are the leading cause of death worldwide, with approximately 64 million of patients suffering from heart attacks that leave behind scarred (infarcted) tissue resulting in impaired heart function. Our team is developing a piezoelectric implant that harnesses the heart’s natural mechanical motion to generate electrical stimulation, promoting cardiac tissue regeneration. Using biocompatible piezoelectric film and in vitro testing, we aim to demonstrate that our implant can use strain induced voltage to promote cardiomyocyte proliferation and restore function in infarcted heart regions without external power sources.

Team:Kellan Yoshikawa [ME], Lin Lee [MSE], Nikolai Seva [ME], Preston Dankwah [ME], Savita Pereira [ME]
Advisor(s): Grace O’Connell
Project ID: 14

Prescribed orthotics offer significant benefits such as arch support, corrected foot posture, and long-term comfort. However, the current dress shoe market prioritizes fashion over support, and shoes are often incompatible with orthotics, resulting in a 31% abandonment rate of inserts. Our shoe bed technology eliminates the need for separate inserts by using CAD to parametrically model a shoe bed personalized to each user’s foot anatomy, allowing users to wear their favorite dress shoes. This breakthrough brings biomechanics into fashion, provides comfort and accessibility, and ultimately promotes long-term maintenance of foot health.

Team: Shereen Aissi [NE], Damanpreet Bhattal [BioE], Isabella Bredwell [NE]
Advisor(s): Lee Bernstein, Andrew S. Voyles
Project ID: 34

Actinium-225 is a promising medical isotope for targeted alpha-particle therapy, but its scarcity limits clinical use. Our team is developing a high-power target design to increase Ac-225 production using hospital-grade cyclotrons originally used for positron emission tomography. Proton bombardment of a beryllium target generates neutrons for the Ra-226(n,2n)Ra-225 reaction to produce Ac-225 without unwanted byproducts. To optimize target durability, COMSOL Multiphysics software is used for determining the critical heat flux, while Monte Carlo N-Particle simulations are performed to maximize neutron flux. This scalable, cost-effective approach aims to alleviate the global Ac-225 shortage and improve cancer treatment accessibility.

Team: Yuxin Ye [ME], Sebastien Guerif [ME], Ki-Hoon Lee [ME], Xavier Johnson [ME] , Sarah Nwakudu [ME]
Advisor(s): Victor Detavernier, Bryan Tran
Project ID: 50

Traditional treadmills support linear movement but fail to fully replicate natural locomotion, limiting their effectiveness in rehabilitation and immersive VR experiences. Omnidirectional treadmills offer 360° full-range motion, however existing models suffer from high cost and large footprints and often cause motion sickness in users due to unnatural movements. Our team with Blue Goji will create a unique take on the omnidirectional treadmill that seeks to improve on these limitations by utilizing a non-powered resistive surface comprised of free spinning balls and a hybrid braking system that employs both friction and magnetic mechanisms. This treadmill aims to allow users to move in any direction while maintaining a natural gait.

Team: Luai Abuelsamen [ME], Swati Priyadarshini [ME], Harsh Rana [ME], Wenhan Tang [ME], Ho-Wei Lu [ME]
Advisor(s): Gabriel Gomes
Project ID: 154

We’re developing a fast, low-cost motion planning system for industrial robots that ensures smooth, collision-free movement. By combining NVIDIA’s Jetson (a compact GPU computer) with cuRobo (an advanced motion planning library), we’re able to compute optimal robot paths in real-time—right at the edge, without relying on cloud servers. This makes automation more accessible and affordable for manufacturers. In partnership with Vention, a leader in plug-and-play factory automation, our solution helps bring high-performance robotics to factories of all sizes.

Team: Olvia Bloukos [ME], Richard Butcher [ME], Rabhat Chaiprapa [ME], Adam Reagen [ME]
Advisor(s): Tarek Zohdi, Scott Ziegler
Project ID: 90

Traditionally, coordinating a satellite launch into space requires years of mission planning. Space Kinetic has developed an in-space launch station, the Longbow, that will project a capsule into a desired orbit, reducing overall mission time. We have designed the capsules that will be on standby in the Longbow, which can be deployed instantaneously for orbital missions.

Team: Jeremy Chen [ME], Junyu Lu [ME], Max Freeman [ME], Eric Pham [ME], Yongye Hu [ME]
Advisor(s): Francesco Borrelli
Project ID: 28

Exploring hazardous environments is a complex task. It is dangerous and requires the ability to adapt to varied environments, for instance, by avoiding obstacles. While robots like drones and rovers help reduce risk, they are often limited to a single mode of movement, sometimes requiring the use of multiple platforms to complete a single objective. Our project seeks to solve this challenge by developing the L.I.F.T. Platform, a hybrid system capable of both driving and flying. This dual-mode capability enables seamless navigation across varied terrain, with potential applications in search-and-rescue, last-mile delivery, infrastructure inspection, and military reconnaissance.

Team: Vlad Rosca [ME], Qiyuan Liu [ME], Yunhao Liang [IEOR], Xinran Yang [IEOR], Weixing Guo [ME]
Advisor(s): Mark Mueller
Project ID: 108

In the U.S. alone, it is estimated that crop losses from plant pathogens cost up to $21 billion annually. Our team is working on a fully autonomous drone model to retrieve soil and crop data to help the agricultural station predict and respond in real time. We employ a built-in path planning algorithm alongside a local trajectory planner to optimize navigation. A Wi-Fi transmission module ensures seamless data transfer of retrieved soil information to our drone. Finally, we integrate all components into the robot operating system to facilitate smooth operation and seamless module collaboration.

Team: Penelope Kim [ME], Andy Yu [ME], Kush Dasadia [ME], Pierre Louis Soulie [ME]
Advisor(s): Francesco Borelli
Project ID: 29

Terrestrial vehicles often struggle with hazardous or uneven terrain, where fixed configurations limit mobility and adaptability. Our project, the Morphable Robot Vehicle for Terrain Adaptability (MRV), introduces a user-configurable chassis with a controller that responds to manual, tool-free adjustments, enabling more effective traversal across varied environments.

Team: Wendy Cheng [CEE], Gustav Martin Friede [ME], Jonathan Goenadibrata [ME], Jipeng Hu [ME], Fernando Pluma [ME], Haoyang Zhou [ME]
Advisor(s): Gabriel Gomes
Project ID: 74

In an event of a disaster, sites often can only be safely entered by first responders many hours or even days later, time that is essential for the survival of victims. We are developing a proof of concept for an automated search and rescue system that can be immediately deployed in post-disaster areas and increases the rate of survival. The solution is a software system integrated in a ROS2-based rover, that uses mapping, exploration and object detection algorithms, to autonomously explore and map a simulated environment and accurately identify and locate survivors.

Team: Tze Yat Max Lee [ME], Ethan Champion [ME], Kendall Cooney [ME], Heather Hartman [ME], Juliana Kew [ME], Ziang Xiang [ME]
Advisor(s): Michael Gollner, Kelly Clevenson, Wuquan Cui
Project ID: 69

Wildfires have devastating effects on environmental and human health. To preemptively combat wildfire spread, the truck-sized BurnBot creates fire breaks by driving over and incinerating fire-prone brush with onboard torches. Our team is building a mechanical detection system to detect potentially hazardous obstacles and determine if BurnBot can pass over them without damaging these torches. Automating this detection process reduces the total number of operators and furthers the goal of making BurnBot fully autonomous.

Team: Bryan Medina [ME], Isaac Krieger [ME], Lino Le Van [EECS]
Advisor(s): Tarek Zohdi
Project ID: 89

As we stand on the precipice of a new space age, the need for rapid situational awareness capabilities in an increasingly dynamic space environment becomes of the utmost importance. To this end, venture-backed start-up Space Kinetic is developing a new technology dubbed “The Longbow”: an electromechanical system that integrates onto a host satellite and deploys a cache of smaller assets without needing to fire a thruster. Our capstone project aims to develop a parameterizable mission plan for this technology, including launch vehicle selection, orbit determination, event timelines, sensitivity analysis, and end-of-life operations.

Team: Jan Dustin Tengdyantono [ME], Kieran Pereira [ME], Mariette Peutz [BIOE], Michael Han [MSE]
Advisor(s): Reza Alam
Project ID: 1

The shipping industry incurs an additional $45 billion annually in fuel costs due to the energy required to overcome waves during voyages. Our objective is to engineer an array of autonomous sailboats that collect oceanic data, such as tidal forces, across large areas. With this data, the algorithm can map oceanic forces and optimize fuel-efficient shipping routes—potentially saving shipping companies up to 30% in fuel costs. Additional use cases are being explored, such as supporting environmental conservation efforts using temperature, salinity, and pH indicators.

Team: Olivia Cassone [ME], Inès Feret [ME], Hursh Singh [ME], Xinyu Wan [ME], Zhen Zhao [ME]
Advisor(s): Feihuan Qu
Project ID: 18

Xper is a tele-operated robotic system designed to enhance precision in interventional radiology procedures. By integrating real-time CT fluoroscopy imaging with a robotic arm and needle guidance system, Xper enables safer, more accurate needle insertions during surgeries like tumor ablations and biopsies. The system reduces radiation exposure and supports remote operation, offering a smarter, more efficient future for minimally invasive care.

Team: Stephen Lee [EECS], Shiran Yuan [EECS], Brayon Lordianto [EECS], May Liu [EECS], Junming Chen[EECS]
Advisor(s): Avideh Zakhor
Project ID: 32

In our increasingly digital world, creating digital environments has emerged as a critical tool for simulation, virtual collaboration, and design. Our team has developed a pipeline for quickly, digitally reconstructing physical spaces using drones at low-cost. By leveraging 360 degree cameras and drones to easily collect data, 3D gaussian splatting has enabled our team to efficiently reconstruct photorealistic indoor scenes for these applications.