2026 ECE Design and Innovation Day

Project Description: Electroencephalography (EEG) is widely used to monitor brain activity and diagnose neurological conditions, but traditional systems are bulky, expensive, and difficult to use outside of clinical environments. This makes long-term monitoring inconvenient and often inaccessible, especially for patients in everyday or remote settings. Our project addresses this by creating a compact, easy-to-use EEG device designed for continuous brain monitoring in daily life. The device is small, lightweight, and comfortable to wear, allowing users to go about their normal routines while collecting data. It connects wirelessly to a simple interface where users can view their data in real time or save it for later review. Multiple devices can be used simultaneously for more comprehensive monitoring, making brain monitoring more practical, accessible, and user-friendly.

A major challenge in this project was delivering reliable performance in a very small device while keeping it safe and easy to use. Gold-plated electrodes were incorporated to ensure consistent contact with the skin for signal collection, and pogo pins were used to provide reliable internal connections within the compact design. We then addressed power delivery by integrating a compact battery with wireless charging, allowing the device to be placed on any standard Qi charger without cables. The system manages charging automatically to ensure safe, efficient operation, making it simple to recharge and reuse. 

Hearing loss affects over 1.5 billion people globally, with that number continuing to rise. Traditional methods of hearing assessment, such as audiometry and tympanometry, require specialized clinical equipment, which can limit accessibility and increase cost. iHear2++ addresses this by providing a low-cost, portable, and easy-to-use system to measure acoustic parameters of the ear. This enables the detection of hearing abnormalities while also supporting applications such as improved hearing aid design and even the creation of custom audio filters to enhance headphone listening experiences.

TheiHear2++ plays a chirp covering the full audio spectrum (20 Hz–20 kHz) through a 3D-printed tube inserted into the ear. Five microphones, positioned along the tube at precise locations, capture the reflections of this chirp. Our hardware then transfers the five-channel signals as PDM data to a host PC containing a specialized GUI. Digital signal processing is used to reformat and reconstruct the data, after which a reflectance coefficient algorithm is applied. The resulting signals are displayed visually in real time, enabling immediate interpretation and analysis. This data can then be used by doctors and audio engineers to develop specialized hearing devices and improve audio systems.

Please feel free to contact us with any further questions.

Lavis Chenlavisc@student.ubc.ca Sophia Cockram scockram@student.ubc.ca, Michael Yan myan13@student.ubc.ca Xianyao Li (xianyao3@student.ubc.ca), Frances Chen(frs0918@student.ubc.ca)

Project Client: Frostad Research Lab at UBC 

Project Description:

Our project focuses on developing a refined, integrated prototype of an Interfacial Dilational Rheometer (IDR) to enable accurate measurement of fluid interface properties for biomedical and industrial applications. Existing systems lack proper thermal control, structural integration, and reliability, limiting their effectiveness under realistic operating conditions.

We have designed a compact, modular aluminum enclosure that improves structural integrity, accessibility, and system organization. To meet strict performance requirements, we implemented a layered insulation architecture that minimizes heat loss, reduces condensation, and enables stable operation at elevated temperatures.

We have developed a low-noise power distribution and regulation system, including AC integration, DC conversion, and filtered voltage rails to ensure reliable sensor measurements and safe operation.

By integrating mechanical, thermal, and electrical subsystems into a single cohesive design, our solution enhances measurement accuracy, system efficiency, and usability, supporting real-world deployment in biomedical research and product development.

Contact Information:

1. Akshat Ranjan Email: akshatranjan@ymail.com, akshat1r@student.ubc.ca

2. Ethan Andre Email: ethanandre@shaw.ca, eandre04@student.ubc.ca

3. Hrishik jain Email: hrishikjain2003@gmail.com, hrishik3@student.ubc.ca

4. Kunal Lugani Email: kunal.lugani1@gmail.com

5. Saksham Mahajan Email: mahajansaksham77@gmail.com

Purpose of your project: What problem is it solving and for whom?

This project addresses the risk of injury from unexpected falls, particularly for vulnerable populations such as the elderly. Falls occur rapidly and often result in serious health consequences due to a lack of protective response time. The solution is an anti-fall airbag vest module that detects imminent falls and deploys protection before impact. This improves user safety, reduces injury severity, and supports independent living.

Major design contribution: What is the technical challenge that you resolved?

The main technical challenge was achieving high detection accuracy while maintaining a compact and modular hardware system. A real-time detection algorithm was utilized and trained to reliably distinguish actual falls from daily activities with high sensitivity and low false positives. In parallel, a modular embedded hardware platform was designed to integrate sensing, processing, and actuation. This allows the system to interface with different deployment mechanisms and remain adaptable for future improvements.

Your contact information if you would like employers to get in touch with you.

jg832@student.ubc.ca

Miniature Ultrasound Transducer

Conventional ultrasound machines are bulky, expensive, and confined to hospitals, making them inaccessible for routine home-based monitoring. Our team is developing a miniature, wireless ultrasound imaging device that enables patients to receive affordable ultrasound scans at home using their personal computer.

Building on a previous capstone prototype, we redesigned the system around a multi-element transducer array enabling B-mode imaging and Doppler blood-flow measurement. The system consists of custom analog front-end and high-voltage pulser PCBs used to transmit and receive ultrasound signals. A central FPGA preprocesses the acquired data and transmits it over WiFi to a laptop, which completes the image reconstruction and display. The hardware is organized as a modular PCB system designed to fit within a compact handheld enclosure. The device is battery-powered, targeting over three hours of continuous operation.

Our major technical contributions include an FPGA preprocessing pipeline that eliminates the need for a dedicated GPU on the host computer, a validated wireless data link capable of real-time frame rates, and a scalable multi-channel architecture that can accommodate different transducer configurations through a standardized interface.

Client: NeuroPrior AI

Team: 

• Harnoor Saigal: hargss@gmail.com

• Isha Shukla: ishashukla2000@gmail.com

• Sidharth Sudhir: sidharth.sudhir2003@gmail.com

• Yousaf Nazari: yousafnazari5@gmail.com

• Yuhan Qiu: qiuyuhan66@gmail.com

Supervisor: Leo Stocco

Stroke rehabilitation in Canada faces a critical gap, particularly for patients who have been discharged from hospital but still require ongoing care. Limited access to outpatient physiotherapy, especially in rural communities, leaves many patients without the consistent treatment needed to regain motor function. Our project targets one of the most important aspects of recovery: restoring movement in the hands through regular, assisted rehabilitative exercises.

Our solution is a wearable assistive glove that performs therapeutic hand exercises on behalf of the patient, eliminating the need for a physiotherapist to be present. This allows stroke survivors to access essential rehabilitative care from home, removing the financial, geographic, and logistical barriers that often interrupt recovery.

The key technical challenge we resolved was building a glove that uses heat-activated artificial muscles to mimic natural human movement, while keeping the device lightweight, non-bulky, and fully washable. We achieved this by constructing the glove entirely from fabric and integrating conductive hook and loop fasteners and thread, allowing all electrical components to be detached before careful cleaning. The arrangement of the actuators was also carefully designed to replicate the mechanics of real hand movement effectively.

Contact information:

rubayo@student.ubc.ca

tyy@student.ubc.ca

at7269@student.ubc.ca 

vidunikankan@gmail.com

idaran@student.ubc.ca

Characterizing the Initial Orthostatic Response using a Multi-function Fingertip Pulse Oximeter (MOx)

Project Client: VitalSines Inc.

Project Description: Orthostatic hypotension is difficult to detect using traditional cuff-based measurements, as they cannot capture rapid blood pressure changes during posture transitions. This limits effective assessment for healthcare professionals monitoring patients.

Our project develops a mobile system that connects to a fingertip oximeter to collect ECG and PPG signals via Bluetooth. The app processes these signals in real time to estimate continuous blood pressure and guide orthostatic testing.

A key technical challenge was ensuring reliable, low-latency BLE data streaming while maintaining synchronization between high-frequency physiological signals. We addressed this through a real-time processing pipeline and robust connection handling.

Liam Clawson-Honeyman liamclawsonhoneyman@gmail.com; Brandon Cheong brandontcheong@gmail.com; Xintong Pan panxintong2003@gmail.com;  Aakaash Senthilkumar aakaashtech@gmail.com; Shitong Zou shitong.zou@outlook.com

Purpose of the project:

Traditional vital sign monitoring equipment is not allowed in hospital seclusion rooms due to their wires posing a self harm risk for occupants. These occupants are primarily patients suffering from psychiatric disorders. As a result, these patients are at a disproportionate risk of having a medical emergency as they are often sedated or prescribed medication that increases the likelihood of cardiac or respiratory arrest. The high risk nature of being secluded requires constant monitoring by medical personnel every 15 minutes to ensure that patients are not in a life-threatening state.

Additionally, vital sign monitoring wearable devices, such as the electrocardiogram (ECG) or photoplethysmogram (PPG), can cause skin irritation and extreme fear in patients already in a fragile headspace. These devices may also be misused, as patients can exploit cables to harm the medical staff on a check-in or for self harm. Manual checks from medical staff increase labor demands and have inherent observation gaps where the patient is at risk of going unnoticed in case of an emergency. 

Our goal is to provide a contactless vital sign monitoring system capable of extracting heart rate and respiratory rate from a sleeping patient and alerting medical staff in the case of a heart rate abnormality, respiratory rate abnormality, or if a fall is detected. This is done through the application of mmWave FMCW radar for vital sign extraction, a thermal sensor for patient presence detection, AI models such as a convolutional neural network and an autoencoder, and finally an event-driven client-server software architecture for device configuration and emergency alerting.

Major design contribution:

The core design challenge was the extraction of heart rate and respiratory rate from Frequency Modulated Continuous Wave (FMCW) radar reflections in real time. Chest movements from respiration displacement and heartbeat are extremely subtle and are often obscured by motion artifacts and noise. To resolve this issue, we developed a custom signal processing pipeline that uses multiple state-of-the-art algorithms to clean through the noise and motion artifacts to produce an optimal vital sign estimation. The system is able to achieve heart rate estimations within 3 beats per minute (BPM) mean absolute error of a contact-based reference heart rate sensor (PPG) under controlled and ideal conditions up to over 1m away from a patient. Similarly, respiratory rate estimates fall within 2 respirations per minute (RPM) mean absolute error. For fall detection, the system uses the same FMCW radar hardware and configuration to convert raw ADC data into range-Doppler maps. It organizes the radar samples into frames, removes static clutter, and applies signal windowing to reduce noise and improve clarity. As for the system itself, it is built on a central Node.js server with multiple long-lived python subprocesses yielding a system that handles data in real time and displays this on a React web dashboard for delivering nearly instantaneous alerts upon detecting a fall and vital sign abnormalities to medical staff.

Yuheng Wu: wuyuheng0525@gmail.com

Justin Kim: justinkim0345@gmail.com

Luc McDonald: lucrwmcdnld@gmail.com | 604-762-5596

Lakshya Saroha: lakshyasaroha43@gmail.com | 604-861-0020

Louie Tang: louietang2013@gmail.com | 778-323-8985

With Small Modular Reactors (SMRs) emerging as a new energy technology, Fluor Corporation aims to evaluate their feasibility alongside other energy sources such as hydro, wind, and solar for remote mining applications. In early-stage project planning, engineers must quickly assess potential power solutions with limited information, making it challenging to compare options in terms of energy generation, cost, and feasibility.

This project expands Fluor’s existing Power Utility Renewable Energy Simulator (PURE-SIM), a decision-support tool that automates feasibility analysis for various energy systems. Our team developed two major additions: an SMR module and a comparison tool. The SMR module models energy generation, economic analysis, and regulatory feasibility for SMRs, while the comparison tool evaluates SMRs against one another and other energy sources to rank optimal configurations for a given load profile.

This solution streamlines the evaluation process, reduces manual effort, and enables faster, more consistent, data-driven decision-making for early-stage project planning.

Ash Murthy: ashrmurthy@gmail.com 

Doyoung An: Doyoung175@gmail.com

Peter Woolsey: peterowen.woolsey@gmail.com

Yuqian Song: yuqiansong85@gmail.com

Tera Minato: minato.tera@gmail.com

This project presents the design and techno-economic assessment of a 2 MWh lithium-ion Battery Energy Storage System (BESS) to support reliable and sustainable energy in remote Canadian communities. Many of these communities depend on diesel generation, which is costly, logistically challenging, and environmentally impactful. The proposed system reduces diesel reliance by enabling the integration of renewable energy and providing stable, dispatchable power.

A primary design challenge was ensuring safe and reliable operation in extreme cold climates down to -50 °C. The system includes a thermally optimized enclosure using a modified shipping container for the battery system and a separate enclosure for power conversion equipment. HVAC and heating systems were carefully sized to maintain operating temperatures, while electrical designs addressed power distribution, protection, and communication systems in compliance with Canadian standards.

In addition to technical design, a techno-economic assessment was conducted to compare the system against conventional diesel generation. Results indicate strong potential for improved reliability, reduced operating costs, and lower environmental impact for off-grid applications.

Project Client: Electrical and Computer Engineering Department, UBC

The aim of our project is to prototype an underwater acoustic energy harvesting system using a piezoelectric transducer to support long-term operation of Internet of Underwater Things (IoUT) devices. Currently, these systems use batteries that require frequent replacement, leading to high maintenance costs.

Our project addresses this challenge by developing a system that converts ambient sound into electrical energy. Low-frequency acoustic waves are input to a piezoelectric transducer, which generates a small electrical signal. This signal is then passed through a rectification stage, boosted to a usable voltage, and stored for later use.

Our design allows for self-powering IoUT devices, lowering maintenance requirements and improving system sustainability.

Contact Information: Julia Wadey juliawadey2003@gmail.com; Jerry Fu jerryfu6158@gmail.com; Joshua Jacob John    joshuajacobj7@gmail.com; Joshua Xu joshua.x3456@gmail.com; Weiming (Jeff) Zeng wzeng04@gmail.com

Project Description: 

One of the biggest challenges in maintaining data center reliability is the early detection of mechanical faults in cooling infrastructure before they escalate into thermal failures or unplanned downtime. Current monitoring workflows are largely reactive, triggering alerts only after a cooling problem has already manifested as a temperature anomaly. Our project aims to develop an intelligent fault detection system that identifies mechanical faults at the cooling fan itself, in real time and without dependence on downstream thermal indicators.

To achieve this, we will collect vibration data directly from cooling fans using accelerometer sensors and apply machine learning classification methods to analyze the resulting signals. Based on the patterns learned from this data, our system will distinguish between normal fan operation and representative mechanical fault conditions. The trained model will be deployed into a live inference pipeline, surfacing fault labels and alerts through an operator-facing dashboard. This approach ensures the system can autonomously detect and classify cooling faults in real time under varying operating conditions, enabling a shift from reactive maintenance to proactive fault awareness without reliance on manual inspection or temperature-based monitoring alone.

Bryan Tanady (tanadybryan@gmail.com); Kevin Chen (kevin.chen.xiv14@gmail.com); Yik Hin Matthew Lau (matthew2004728@outlook.com); Kei Matsui (keimomo@student.ubc.ca); Yujun Lai (zaneyujunlai@outlook.com)

Our project aims to solves the problem of low-fault current protection for microgrids for BC Hydro. With the rise of renewable energies and grid-scale batteries there is an increasing penetration of inverter-based resources in modern grids, which poses a new problem for power grid protection. Normally, with conventional sources of energy, when a fault occurs (i.e. a disturbance in flow of power from occurrences such as a tree falling on a power line), a large amount of current, roughly 9 to 10 times the rated amount, will flow into the fault. Conventional protection systems utilize simple overcurrent protection to detect this large current flow to protect grid assets in case of faults. However, with inverter-based resources, the internal electronics limit the available max fault current to roughly 1.5 to 2 times the rated current, meaning that the current is not high enough to be detected by conventional protection equipment. We solved this issue by designing a new protection scheme based on differential protection, comparing currents entering and leaving two separate points of a power line to detect a fault within. Since our proposed system utilizes differences in current for fault detection rather than the magnitude of current, it provides sensitive yet fast responding protection that can be used for any level of fault current.

Purpose

AquaSentinel is an end-to-end system designed to help communities in low-connectivity regions access safe, potable water. It continuously monitors key water quality parameters, such as pH, dissolved oxygen, and turbidity, and automatically alerts local operators and residents when conditions become unsafe.

Design Contribution

The system integrates solar-powered sensor stations that transmit readings over a LoRa-based wireless network to a central gateway. An on-device machine learning model classifies water quality in real time, while a cloud-connected dashboard displays both live and historical data trends. When thresholds are breached, the platform sends SMS notifications to ensure rapid awareness and response. In areas where internet coverage is unreliable, the gateway can host the dashboard locally and maintain ongoing data collection, ensuring communities remain informed even during network outages. AquaSentinel’s goal is to provide early detection of water contamination and empower proactive action before public health is at risk.

Member Contact

Anu Ponnusamy: anumithrap@gmail.com

Alvin Shon: ssw7567@gmail.com 

Justin Ng: ngjustin2002@gmail.com 

Reanna Wong: reanna.s.wong@gmail.com 

Yizhou Zhou: yizhou11182@gmail.com 

Project Title: AirWatch – Distributed Air Quality Monitoring System

AirWatch is a distributed air quality monitoring system designed to provide accessible, real-time environmental data for underserved communities. Poor air quality is often difficult to detect without specialized equipment, yet it has direct impacts on human health, especially in urban and industrial areas. Our project aims to make air quality monitoring more affordable, scalable, and easy to deploy.

The system consists of multiple sensor nodes that measure key indicators of air pollution, including particulate matter (PM2.5), carbon dioxide (CO₂), ozone (O₃), carbon monoxide (CO), ammonia (NH₃), nitrogen dioxide (NO₂), temperature, and humidity. These nodes communicate wirelessly and form a resilient network, allowing data to be collected even if individual nodes fail. The collected data is transmitted to a central platform for visualization and analysis.

A major technical challenge we addressed is integrating multiple environmental sensors with different communication protocols into a stable embedded system, while maintaining low power consumption for long-term deployment. We also designed a flexible communication architecture that supports both short-range mesh networking and long-range data transmission.

AirWatch demonstrates how low-cost embedded systems and intelligent sensing can be combined to build scalable environmental monitoring solutions. This project is particularly relevant for smart cities, public health monitoring, and environmental awareness initiatives.

Contact: 

Jenny Gao: 0916gaoyang@gmail.com

Eric Wang: 1330848738w@gmail.com

Zhe Wang: zhe.wang@airwatch.cc

Jingrong Tian: jingrongvivian@gmail.com

Jason Zhong: jzhong.ca@gmail.com

BC Forest Sentinel addresses the need for ultra-early wildfire detection in remote areas where traditional satellite and camera systems are often too slow, and commercial IoT alternatives are prohibitively expensive. Designed to protect BC forests and at-risk communities, this open-source, low-cost system detects fires during the smoldering phase—sensing temperature, gases, and smoke—before visible flames even appear.

Specifically, we developed a self-healing, solar-powered mesh network using ESP32 microcontrollers and Edge AI. The system utilizes TinyML running locally on the sensor nodes to accurately distinguish real fire signatures from false positives (like vehicle exhaust). Data is then reliably routed through the forest via multi-hop communication to a gateway node, which uses the cellular network to trigger real-time dashboard alerts.

Amir Farah (amirsfarah@gmail.com)

Isabella Clyde (isabella.hk.clyde@gmail.com)

Engy Sadik (engymaged98.es@gmail.com)

James Price (pjames266@gmail.com)

Nakul Dharan (nakulgd@gmail.com)

Purpose of the Project

Validating satellite communication links before launch is challenging, expensive, and limited in realism. Once deployed, a loss of communication can render a satellite unusable with little opportunity for recovery, resulting in significant financial and material loss. While low-power protocols such as LoRa show promise for satellite IoT applications, their performance in Earth-space environments is not yet well characterized, and there is currently no convenient, repeatable, accessible ground-based method to emulate realistic Earth-space channel effects with full link observability.

The DCE solves this for CubeSat developers, satellite IoT researchers, and university research labs by providing a controlled, over-the-air (OTA) testbed that reproduces key propagation effects (Doppler shift, free-space path loss, fading, and delay) in a laboratory setting using existing RSL infrastructure.

Major Design Contribution

The core technical challenge was reproducing time-varying satellite channel effects in real time using integrated RF hardware and software, rather than relying on static worst-case assumptions.

Two sub-problems had to be resolved:

1. Real-Time Attenuation (Path Loss) A combination of fixed attenuators established a baseline loss and compensated for RF signal chain limitations, while programmable attenuators were updated at discrete time intervals according to a MATLAB orbit simulation profile, allowing the physical RF path loss to track the modeled satellite pass in real time via automated SCPI commands over serial.

2. Real-Time Doppler Shift The SDR first downconverts the received analog RF signal to a digital baseband representation, allowing channel effects to be applied with high precision through DSP blocks. The time-varying Doppler profile is then applied as a discrete frequency offset, and the modified baseband signal is upconverted back to analog RF output.

The underlying orbital dynamics driving both profiles are computed using the MATLAB Satellite Communications Toolbox, which simulates a CubeSat using standard orbital parameters such as semi-major axis, eccentricity, and inclination, and computes the satellite’s motion relative to a ground station to produce time-varying profiles of Doppler shift, free-space path loss, atmospheric gas attenuation, rain attenuation, and more.

Collectively, this resolved the challenge of bridging a physics-based orbital model to real-time, hardware-in-the-loop RF impairment injection, inside a GTEM cell OTA environment, at a fraction of the cost of in-orbit testing.

Contact info: 

Marcus Cheng – hello@marcusc.me

Connor Froese – connorjfroese@protonmail.com

Brandon Seo – seoobrandon@gmail.com

Hassan Haider – Hassanmc123@hotmail.com

Douglas Zhu – dzhu2003@student.ubc.ca

Our project, the Memory Subsystem Power Estimator, helps engineers quickly estimate the power consumption of DDR5 memory subsystems in server systems. As AI and high-performance computing continue to grow, memory power has become an important part of overall system energy use. However, existing simulation tools are often too slow and complex for fast design decisions.

Our project addresses this by building a web-based platform that provides high-level power estimates using workload statistics and memory specifications. A key technical challenge was translating detailed DRAM power behaviour into a model that is both accurate and efficient, while presenting the results in a clear, interactive way. This tool can help system designers compare configurations, understand power breakdowns, and make better deployment decisions earlier in the design process.

Songli Du: du.songli@gmail.com
Kenn Du: kenndu6243@gmail.com
Lucas Song: lucas02.song@gmail.com
Riley Zhang: ryyizg8@gmail.com
Sonya Zhao: sonyazhaox@gmail.com

Project Client: 

UBC ECE SoC Shekhar Group

Project Description:

In optical systems such as precision LiDAR or coherent optical communication systems, laser stability makes or breaks the system. All lasers suffer from phase noise and frequency drift, which can severely limit their performance in sensitive applications.

To solve this, our project uses the Pound-Drever-Hall (PDH) laser stabilization technique that locks the frequency of a laser to a stable optical reference. By developing feedback sensing and control systems with the optical reference, our system can continuously monitor and adjust the laser’s output in real-time, reducing frequency drift to create a low-linewidth, stable laser.

Developing this system required understanding how to develop electro-optical and mixed-signal systems to detect and amplify modulated optical signals, perform weak signal demodulation, and implement a closed loop digital feedback control system. The result is an effective system that implements the theoretical background into a high-performance, modulator, and reconfigurable engineering solution.

Team Contact Information:

• Alex Lamyin: amlamyin@gmail.com
• Ana Bandari: anabandari@gmail.com
• Eli Rodrigues: erod2662@gmail.com 
• Lochlan Rode: lochlan.rode@outlook.com
• Nolan McCleary: nolancmccleary@gmail.com

Early and reliable ice detection targeted for de-icing critical infrastructures is impeded by the lack of processing sensor data without cloud connectivity. To tackle this challenge, we have collaborated with Okanagan Microelectronics and Gigahertz Applications (OMEGA) Labs to integrate their microwave sensors with our machine learning model to enable robust edge computing. The microwave sensors measures reflection and transmission coefficients (S11, S21) for surface monitoring. However, they lack categorization of these signals under noisy, real-world conditions into bare, water, or ice surfaces; this requires our custom RTL deployed on a field programmable gate array (FPGA).

We have explored several different machine learning algorithms, ranging from a simple random forest to a complicated multi-layered perceptron. Our analyses found that Support Vector Machine (SVM) achieved highest performance when classifying microwave sensor data. We evaluate the algorithm’s performance using both controlled laboratory measurements and synthetically generated noisy datasets to ensure high accuracy and resilience before porting the validated model onto the FPGA for real-time inference.

This project advances OMEGA Lab’s research toward a fully deployable, commercial ice detection product that enhances safety and efficiency across varied industrial applications.

Keywords: machine learning, field programmable gate array, RTL, edge computing, hardware deployment, ice detection, microwave sensors, signal classification

In modern electronic equity markets, a competitive advantage is defined by speed. Trading participants must ingest high-volume market feeds and execute trading decisions in microseconds, where even fractional delays can result in missed opportunities. Traditional software-based trading systems running on CPUs often struggle with operating system overhead and unpredictable scheduling delays, limiting their responsiveness.

Our project explores ultra-low latency computing by designing a hardware-accelerated High-Frequency Trading (HFT) system for our client, Synopsys Inc. By offloading latency-critical tasks to a Field-Programmable Gate Array (FPGA), our system bypasses traditional software bottlenecks to achieve highly deterministic, microsecond-scale performance. This project serves as an educational proof-of-concept demonstrating how advanced hardware design can push the boundaries of performance in software design and financial technology. ​

Do Hyun An linkedin.com/in/dohyunan/; Harris Mai linkedin.com/in/harristmai/; Joe Li linkedin.com/in/joeli-/; John Song linkedin.com/in/johnsong12/; Sofiya Spolitak linkedin.com/in/sofiya-spolitak/

In a world filled with blackbox technology, the SYCL Badge V2 stands as a reminder that software can be simple, transparent, and built with genuine care; core ideas behind the Software You Can Love (SYCL) conference. This compact board is an evolution of the previous design, featuring a vibrant high-resolution display, tactile buttons, lights, and built-in sensors, all powered by a custom-built operating system written entirely in the Zig programming language.

While it looks like a retro handheld gaming device, it is a powerful tool built specifically for SYCL conference attendees to explore and build in Zig, a modern language that prioritizes clarity and performance. The badge moves the development experience off the laptop screen and directly into your hands. Users can create programs that come to life instantly on the device, from vibrant animations and interactive tools to community-made games.

The V2 Badge introduces significant upgrades compared to its predecessor. Key hardware changes include a redesigned board with an upgraded microcontroller, dynamic power sourcing, and complementary peripherals, enabling the badge to manage power efficiently and run reliably with any program. On the firmware side, the added operating system includes a Hardware Abstraction Layer, debug console, file management system, and pre-built functions to improve the user development experience. Additionally, an online web emulator enables conference attendees to virtually test their code without requiring the physical badge.

For more information on the SYCL conference, visit https://softwareyoucanlove.ca/

For further information or to discuss potential collaborations, please reach out to Vanessa Chu (vanessachu333@gmail.com), Ivan Cheung (ivancheung455@gmail.com), Katie Goncalves (katie.goncalves@gmail.com), Sarah Oskuei (sarahoskuei@gmail.com), and Eric Tannant (etannant@gmail.com).

Modern data centers rely on extremely fast chip-to-chip communication, but operating at 224 Gbps creates major power and heat challenges inside the receiver. Our project explores how to make these high-speed data links more energy efficient. To keep communication accurate, receivers use signal-correction hardware to clean up distortion and prevent errors, but this process can consume significant power. The goal of our project is to reduce that overhead by identifying when parts of the correction hardware are not needed for a given connection and turning them off when possible. We developed and evaluated algorithms to do this automatically, with the aim of lowering energy use and cooling demands while maintaining reliable communication.

1. Purpose of your project

Our project addresses the inefficiency and inconsistency of manual device probing for Professor Lukas Chrostowski at the UBC Photonics Lab. While the lab has fabricated a chip containing 66 PIN carrier-injection modulators, the extensive time and data volume required for manual measurement have prevented a full characterization campaign. By designing and building an automated measurement machine, the project provides the lab with the tools necessary to achieve a tenfold increase in testing throughput, standardizing data collection and enabling predictive modeling for future photonic designs.

2. Major design contribution

The core technical challenge involves the integration of a hardware-software automation system capable of unattended operation. Our team resolved this by developing scripted workflows that coordinate instrument control for voltage, wavelength, and temperature with automated motorized stage positioning for wafer scale testing. Furthermore, our project bridges experimental results with predictive analysis by creating a ML framework trained on a combination of measurement datasets and existing simulation results to forecast performance for new modulator designs.

Project Descriptions: In the modern age of the food industry, food processing and production relies heavily on automation which is crucial to ensure the food supply chain is up to demand. Our project is partnered with FPS Food Process Solutions Corp, a company focusing on building industrial food processing freezers and equipment. Our goal as a team is to explore and implement new engineering solutions into detecting costly production shutdowns caused by conveyor belt malfunction in industrial freezers. The Industrial Freezer Early Warning System (IFEWS) focuses on identifying two common failures: belt shuffling, where low belt tension causes radial belt movement, and belt flip, where excessive tension causes vertical belt movement.

The Industrial Freezer Early Warning System (IFEWS) project provides a non-invasive solution to detect irregularities in the belt’s operation in real-time, thus to provide the site engineers and operators an early warning before a catastrophic failure occurs, which is a major limitation of the current system that relies on physical stoppers and estimating tension sensors.

Major design contribution: The main technical challenge and contribution our team focused on for the Industrial Freezer Early Warning System (IFEWS) is a 3 Method Belt Flip and Shuffle monitoring system designed to operate and integrate within the extreme conditions of an industrial freezer, comprising of both Vision and Belt Data Detection algorithms. The primary technical challenge during program development were computing a real-time detection system that meet both food and sanitation requirements, while still be able to function under extreme environments such as temperature ranging from -40°C to 80°C while integrating seamlessly with the client’s existing Hardware and Programmable Logic Controller systems.

Right now, thousands of talented athletes go unnoticed, not because they lack skill, but because no one captures their game. We developed a low-cost, autonomous basketball tripod designed to track, record and stream games for athletes at all levels.

The biggest challenge was achieving real-time, smooth tracking under strict hardware constraints. We addressed this by deploying lightweight YOLO-based object detection, paired with optimized embedded control loops to minimize latency and ensure stable motor actuation.

The result is a system that lets athletes and teams stream and record their games effortlessly, making it easier to get recruited, improve performance, and share their talent with the world.

Albert Zhu (zhu.a111@hotmail.com), Arsh Kang (21arsh.kang@gmail.com), Eason Feng (fengguoyi1223@gmail.com), Jerry Li (jiayuli143@gmail.com), Kevin Li (lkevin2003@gmail.com)

Project purpose: The purpose of our project is to create a lightweight and compact UAV to solve the problem of mapping and 3D reconstruction of an indoor space where human reach may be inconvenient or dangerous. The main client for the project is ICON labs, a research group based in Columbia University.

Design contribution: The main technical challenge that we resolved was the miniaturization of the system, from a drone wingspan spanning over 2 feet to a diameter around 7 inches, while keeping the main software and hardware features of the drone.

contact info: danieljo142@gmail.com   aaranp1919@gmail.com  ehmsmith12@gmail.com   will1am6@student.ubc.ca   g.smith.191104@gmail.com

As the space domain becomes increasingly congested and contested, the detection and monitoring of geostationary satellites are critical for maintaining Canada’s security. As the number of satellites and debris in orbit continues to rise, organizations such as MDA Space Ltd. face growing challenges in developing cost-effective and scalable observation
systems to ensure the safety and sustainability of space operations. Furthermore, adversarial proximity operations and the proliferation of large satellite constellations have heightened the risk of collision and interference.

This Capstone team has developed a desktop-sized autonomous micro-optical geostationary observatory (MOGO) that is capable of targeting and imaging geostationary satellites for later analysis and monitoring. A user can send a list of satellites to observe overnight and MOGO will autonomously find and image each satellite. The images MOGO produces allows users to monitor satellites over time, watching for debris strikes or other satellites coming too close. This system is compact and low-cost, increasing accessibility to satellite monitoring.  

Armeen Abedini: armeen.abedini@protonmail.com

Matei Dragomir: matei.a.dragomir@gmail.com

Alex Manak: alexmanak@outlook.com

Nathan Roorda: nathanroorda@gmail.com

Deepak Thiagarajan: deepakroshan73@gmail.com

Currently, robotic arms in assembly lines or in humanoid robotics usually rely on vision- or light-based systems to navigate their environment and reach a given target position. However, this approach is ineffective in environments where a clear line-of-sight isn’t available, for example due to obstacles between the robotic arm and the target. For effective navigation in these cases, the arm would need access to other types of sensors.

The Molecular Mechatronics lab at UBC has developed a new technology for low-cost, large area flexible tactile sensors and is interested in using these sensors to give these robotic arms a sense of touch, similar to humans, so that they can navigate in these environments by detecting when and where contact with an obstacle occurs. To test the feasibility of this application, our client Chrys Morton from the MM lab tasked us with designing and manufacturing a suitable sensor based on this technology, and integrating it into a control system that uses it to help a robot arm reach its target while responding to obstacles it encounters.

We designed our sensor and mounting hardware for use with the wrist link of a Franka Robotics research arm, and developed a reinforcement learning-based control system, for which we created a virtual replica of our sensor and a training environment with randomly positioned obstacles to allow us to train our policy. We also developed a framework to allow us to deploy our policy on the real robot arm using ROS2, and a BLE sensor readout driver to interface with the control system running on a Linux workstation.

Huiyu Chen mussymuxi@gmail.com; Jiayi Chen engcjb@gmail.com; Henry Rovner rovnerhenry@gmail.com; AnthonyJinke Su sujk23279@gmail.com; Alan Xue luo@tianyi.vc; Hanlin Yu yuhanlin29@gmail.com.

Project Description: Aarcomm designs and manufactures industrial wireless remotes that must meet durability standards such as IEC 60068-2-32 and SAE J1455B 4.11.3.1. Currently, Aaromm’s engineering  division lacks a repeatable and reliable way to verify that their products meet these standards. We have agreed to deliver an apparatusthat will facilitate and standardize their drop testing procedure, by dropping remotes of various sizes from 1m, 1.5m, and 2m in six different positions.

Akaash Mahal: akaash.mahal@gmail.com, Will Johnson: willjr03@outlook.com, Yash Bhardwaj: bhardwajyash03@gmail.com Taku Nyamupa: takaug2002@hotmail.com, David Joseph: dbjoseph9@gmail.com 

Client: Foundation of the Energy Collective

Project Description: This project develops a remote monitoring platform for an automated wheatgrass production system, addressing the challenge of supervising crop conditions in environments where frequent on-site access is limited. The goal is to provide operators with continuous visibility into system performance and environmental conditions to support reliable and efficient crop growth.

The system integrates multiple environmental sensors, including temperature, humidity, light intensity, and water flow, with a web-based dashboard that presents both real-time and historical data. Users can visualize trends through interactive graphs, view system status through a schematic interface, and receive alerts when conditions fall outside predefined thresholds. These features enable early detection of issues and support data-driven decision-making for maintaining crop quality.

A key contribution of this project is the design and implementation of a full-stack IoT architecture that connects embedded sensor nodes to a centralized backend and user-friendly frontend. The system ensures reliable data transmission, structured storage, and efficient visualization, while remaining scalable for future expansion.

Tahsin Hasan (tahsinhasan2@gmail.com); Hamza Islam (hamza98@student.ubc.ca); Sam Pan (sampty@student.ubc.ca); Ao Sun (aosun620@gmail.com)

Project Client: Bosch GmbH

Purpose of the Project:This project addresses the challenge of accurate indoor localization in environments where GPS cannot be used. Bosch GmbH is exploring opportunities to expand beyond automotive and consumer products by developing indoor positioning modules. Our team is designing a system that combines Bosch inertial measurement units (IMUs) with ultra-wideband (UWB) technology to provide precise indoor location tracking. This technology could be used to track robots in warehouses, monitor patients in care homes, and improve worker safety in environments such as mines.

Major Design Contribution:The main technical challenge we addressed was improving indoor positioning accuracy by combining data from IMUs and UWB sensors. IMUs provide continuous motion tracking but tend to drift over time, while UWB offers accurate position measurements but can be limited by signal conditions. Our design integrates these technologies to reduce positioning errors and create a more reliable indoor localization system.

Our project, developed with Tern Robotics, explores how unmanned aerial vehicles (UAVs) can do more than just capture video during border and remote-area monitoring missions. Examining large areas such as the Canada-US border is difficult, expensive, and often limited by darkness, distance, and poor connectivity. While long-range UAVs can collect valuable aerial footage, human operators still need to manually interpret what they see, which slows response time and increases workload. To address this, we built an onboard perception prototype that can automatically detect, classify, and track people and vehicles directly on the UAV in real-time.

A major design challenge was making the system work reliably under real deployment conditions, including UAV motion, changing altitudes, harsh weather, and low-light environments, all without relying on cloud computing. We addressed this by combining lightweight AI models like the RF-DETR Nano and ByteTrack, RGB and infrared sensing, and onboard computing on a Jetson Nano microcontroller to create a compact perception system. We also calculate object speed estimations to provide an additional layer of situational awareness. Beyond the onboard system itself, we containerized our training pipeline and migrated it to Google Cloud Vertex AI, creating a reproducible workflow for dataset management, experiment tracking, checkpoint versioning, and large-scale model training. This makes the project easier for Tern Robotics to extend, since they can retrain models, adjust datasets, and test future improvements in a structured cloud environment rather than rebuilding the workflow from scratch.

According to the USDA, farmers lose $150 million in direct costs to crop damage from birds. This financial strain is compounded with a rise in farm theft, with thieves targeting high value crops. While traditional deterrents such as scarecrows, reflective tape, and surveillance cameras are common, they only affect a limited area around them. On large-scale farms, these methods leave gaps in coverage, making it difficult to monitor the entire property effectively.

The project addresses this gap through a scalable and reliable solution centered on an autonomous airship.  Our team has implemented the software component of the solution, while another team has implemented the hardware component. The software processes live camera feed through a custom machine vision model designed to identify birds and people in real time.

Upon detecting a threat, the system initiates an immediate response by playing a deterring sound to ward off pests while simultaneously sending notifications to the user. This end-to-end functionality is managed through an intuitive web dashboard where farms can monitor live video feeds and access a library of recorded clips for analysis. The result is a highly customizable and user-friendly surveillance ecosystem that places total control and real-time information directly into the hands of the farmer.

Purpose of the project

This project addresses the problem of identifying maritime vessels when conventional GPS tracking information is missing, unreliable, or intentionally disabled. This gap is critical for organizations involved in maritime monitoring and enforcement, as unidentified vessels are often linked to illegal, unreported, and unregulated (IUU) fishing, smuggling, and poor working conditions. To help address this, we developed a system that uses satellite imagery to identify vessels based on visual appearance and connect new detections to past sightings. By linking these otherwise isolated observations, the system would provide analysts with another layer of maritime intelligence when self-reported vessel identity data is unavailable.

Major design contribution

The major design challenge of this project was building a reliable vessel reidentification model using low resolution satellite imagery. Unlike typical image datasets in object re-identification problems, satellite data is often low resolution and affected by clouds, making it difficult to isolate meaningful visual characteristics.  On top of this, many vessels have very similar visual characteristics from this distance, making it difficult to distinguish one boat from another with confidence.

Our main technical contribution was the design of a machine learning and data-quality pipeline capable of extracting useful vessel characteristics despite these limitations. To improve reliability, we explored and refined multiple strategies, including background normalization, length normalization, image augmentations, data filtering, different data import approaches, and alternative model architectures. Through this process, we found that a triplet-loss-based approach with a ResNet backbone provided the strongest performance for learning discriminative visual embeddings from noisy and visually similar vessel data. This design allowed the model to better separate similar-looking vessels while making the most of the imperfect imagery available.

Contact us!

Miguel Menard – Miguel.Menard@me.com

Emily Wood – emily.wood1624@gmail.com

Milica Maksimoviic – milicamaksimoviic@gmail.com

Cameron Fletcher – camvfletcher@gmail.com

Daniel Ding – dyixuanding@gmail.com

Project Client: AIgribot

Project Description: One of the key challenges farmers face is the lack of reliable methods for surveillance of their crops. Existing aerial surveillance devices are not capable of operating for long periods of time. Our project aims to solve this problem through the application of Lighter-Than-Air (LTA) miniature airships. One of the applications of this airship is farm surveillance and deterrence, through the external addition of a speaker module, farms can reduce the number of crops destroyed by animals. The airship can also be used for other applications such as surveillance of locations that are difficult for humans to access often for extended periods of time. This product is targeted for individuals who require surveying large areas of land for long periods of time.

Major Design Contribution: The largest design challenge that is solved is the longer period of time our airship can operate for compared to off-the-shelf consumer drones for the same payload. A key challenge we encountered and overcame was connecting our hardware to a web service for users to monitor their airship telemetry and live camera feed. The software and hardware were selected with consideration for modularity and the airship can be customized to connect external peripherals as per the needs of the user such as a speaker and machine vision modules.

Contact

Kenrick Marcell Haditio – kenrickmarcellhaditio@gmail.com
Pratik Pushkarna – pratikpushkarna2004@gmail.com
James Huang – kaiminghuang04@gmail.com
Faaiq Majeed – faaiqmajeed2004@gmail.com