Industry-Sponsored Student Capstone Projects

ClairVoyan is a vision-based Detect-and-Avoid (DAA) framework for small unmanned aerial vehicles (UAVs), developed in partnership with Guide Air Labs. Unlike conventional DAA systems relying on radar or ADS-B transponders, which are too heavy for small platforms, ClairVoyan employs a passive, camera-only pipeline integrating deep learning detection to identify non-cooperative airborne intruders at long range. The system targets compliance with RTCA DO-387 operational performance standards.

Project SHIELD (Smart Hardware Inspection for Early Latency Detection) addresses a critical gap in safety-critical embedded systems: the lack of proactive, on-device sensor failure detection. Sensor degradation in UAVs, autonomous vehicles, and medical devices often goes undetected until functional loss, with potentially catastrophic consequences. To support a predictive TinyML framework, the team designed a custom DAQ PCB around an ESP32-S3, integrating eight sensor modalities including IMU, barometer, microphone, temperature, photodiode, vibration, and current sensors. Data collection spans nominal operation and fault injection, generating a labeled dataset for ML model training.

Harper Engineering Company sought a fresh look at its 35,000-square-foot Renton factory to better understand how parts and materials moved through the facility and where layout or process changes could improve operations. The project sought to develop a digital factory floor map and document current flow from raw materials and purchased components through machining, fabrication, sub-assembly, final assembly, and shipping. The student team examined inventory kitting, work instructions, and how parts and subassemblies were presented to operators, with the goal of identifying opportunities to improve clarity, ergonomics, material handling, and overall efficiency. The team provided Harper with a current-state view of its operation along with a conceptual “golden cell” design and related recommendations for future improvement.

Productivity often declines in “Third Place” settings like commuting or between meetings, where attention is fragmented and context switching is frequent. Mobile professionals, especially executives, may lose ideas, disrupt cognitive flow, and delay follow-ups. Existing tools are fragmented and non-agentic, forcing users to manually connect capture, organization, and action across apps. This project proposes an agentic AI workflow that integrates a lightweight assistant with digital inking and immersive communication, enabling fast capture, structured next steps, and seamless follow-up without leaving the current context.

Hytek Finishes currently uses manual beacons to communicate machine statuses; but these systems only provide basic visual signals and are found ineffective by workers. The existing system also fails to notify all relevant personnel of status changes. This project developed an integrated system featuring four-state industrial beacons and a touchscreen interface that combines visual indication, issue logging, and notifications while withstanding the plant’s harsh environments. Through its scalable design and integration with company servers, the system enables comprehensive issue tracking, enhancing communication, minimizing downtime, and supporting process optimization at Hytek Finishes.

Quantum processors require each circuit’s virtual qubits to be assigned to specific physical qubits. That mapping can significantly influence algorithm performance because device noise, limited qubit availability, and hardware connectivity constraints may affect execution quality. This project aimed to explore an AI-based qubit mapping capability for this combinatorially large optimization problem using an AlphaZero-style reinforcement learning approach. The student team aimed to support identification of qubit assignments that could better align with processor constraints and potentially reduce the noise experienced by a circuit.

This project developed a RAIN RFID based smart storage system for bottle detection and relative fullness estimation. Using commercial RFID readers and passive tags, this team collected RSSI, phase, and frequency response data to analyze how RF signals vary with liquid levels. To improve robustness, they introduced reference tags to mitigate environmental variation and signal drift. Through controlled cabinet experiments, they addressed challenges such as interference and material effects, demonstrating a low cost and scalable solution for multi object sensing and intelligent storage applications.

Test engineers rely on complex computer networks to coordinate tests on products. When a fault occurs in the network, engineers need to be notified so that tests can resume as soon as possible. This project solves the problem of test networks going down for extended periods of time by notifying engineers as soon as an issue is detected. The system continuously monitors and logs the status of the network. Once a fault is detected, engineers are notified via email to check the logs to determine which part of the network has failed, preventing major delays.

This project addressed limitations in a mechanical polishing system that relied on manual, skill-dependent angular adjustments. It aimed to convert that process into a hands-free polishing platform by integrating motorized goniometer stages for tilt and theta adjustment, a motorized z-stage, and a Windows-based graphical interface. The system was intended to support very fine angular changes on the order of less than 0.01 degrees, provide variable sample loading from 20 g to 400 g, and guide operation through an intuitive interface that reduced the need for manual expertise. These results will provide a more reliable and accessible method for angular adjustment, along with a gentle, repeatable touchdown sequence designed to reduce the risk of sample damage during polishing.

The project developed a technoeconomic and experimental evaluation framework for a 100 t/yr nanoparticle manufacturing process. The analysis estimated capital costs, annual operating costs, CO2e emissions, and material cost per kilogram while identifying major cost drivers in the baseline process. Alternative routes, such as reagent substitution or process modification, were evaluated for potential economic advantages and used to guide experimental nanoparticle synthesis. The resulting materials were intended for fabrication and testing in coin or pouch cells at the Washington Clean Energy Testbeds, providing proof-of-concept prototypes and supporting comparison of baseline and alternative manufacturing approaches.

Jetted tubs provide well-known health benefits, with pain management being one of the most commonly cited, and they are a popular noninvasive option in obstetrics because they do not require pharmaceutical intervention or invasive monitoring. However, hospital water systems can carry pathogenic bacteria such as Legionella, and current mitigation methods like periodic hot water flushes temporarily make tubs unavailable for use. This creates a need, in both hospitals and homes, for a low-maintenance way to keep jetted tubs clean with minimal user interaction. This student team focused on identifying cleaning technologies and operating conditions for a jetted tub cleaning system intended to work with minimal user interaction, avoid direct contact between the cleaning solution and the user, limit water use to less than 5 gallons per cleaning cycle, trigger automatically after a bath or after a set time period, and avoid degrading pumps or tub components. They aimed to produce a prototype concept for keeping a jetted tub system clean, along with findings on customer pain points, estimated costs, testing results, and material compatibility recommendations.

The I-5 Lid project addressed a long-standing need to reconnect neighborhoods in downtown Seattle that were divided by the interstate. Building on a broader feasibility study conducted with WSDOT and WSP, this work investigated structural systems that could support a lid over I-5 and help evaluate how new public space, housing, and development might be accommodated above the freeway. The project considered girder concepts for different column-line layouts using structural steel and precast concrete, along with related effects on foundation loading, vehicle clearances, and other site constraints. It also examined how differing elevations at the east and west edges of the lid could affect access across the structure, with particular attention to the complexity of I-5 ramps and potential barriers to accessibility that would need further study. In addition, 3D CAD models of existing conditions and possible lid layouts were created to support continued analysis and future phases of the broader Lid I-5 effort.

This project focused on a Phase 1 effort to support the search for resin formulations that could improve compression performance in carbon fiber reinforced polymer composites for weight-sensitive structures such as aircraft wings. The project focused on aerogel-and-epoxy systems as a possible route to co-continuous crosslinks associated with resin stiffness and toughness. The student team surveyed commercial aerogel products, compositions, manufacturers, and factors affecting nanoparticle crosslink formation, including surface chemistry, particle size, loading, and host material. The project also aimed to document crosslinks in a selected aerogel using microscopy and SEM, and to create a data-compilation method that could combine supplier datasheets and generated test data into a single structured table while identifying duplicate or equivalent property labels.

This project addressed limitations in conventional incentive spirometers, which lack a standard way to monitor patient use and technique and require separate procedures for oxygen or albuterol administration. The work focused on a therapeutic incentive spirometer prototype that combined a mechanical spirometer, digital user interface, and integrated inhaler/administration port in one device. The proposed design reduced device size, improved handle ergonomics, repositioned the visual performance indicator, and added a port for oxygen and albuterol delivery. Its electronic interface was intended to track inhalation targets, detect diaphragm movement, display successful sustained maximal inhalation, count completed breaths, and provide reset, power, reminder, motion-sensing activation, and rechargeable power functions. The prototype was designed to demonstrate the fit, feel, form, and function of a market-ready medical device while supporting future evaluation of a more consolidated process for respiratory therapy and patient compliance tracking.

The project focused on designing and prototyping a compact, solar-powered oceanographic buoy intended to operate an underwater camera and data transfer system for up to a year with minimal batteries. The buoy concept needed to integrate solar panels, onboard batteries, charge control electronics, a control computer, a camera, and a cellular antenna, along with a mounting frame designed with buoyancy foam, anchoring, and deployment and recovery pick points. The design also needed to target practical field use by keeping the buoy under 1 meter in diameter and about 100 pounds so it could be transported and deployed from a small vessel, while accounting for wave exposure, corrosion, and biofouling. Completion of the prototype enabled MarineSitu to evaluate it for long-term deployment and potential commercialization. This capability would enable persistent underwater ecosystem and biodiversity monitoring at marine conservation sites without cabling the buoy to shore for power.