Student News Brief

Senior project spotlight: Steven Cho

Using wifi to improve avalanche rescues

Harvard SEAS student Steven Cho

For a senior capstone project, Steven Cho designed an aluminum detection system for use in avalanche rescues

Engineering Design Projects (ES 100), the capstone course at the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS), challenges seniors to engineer a creative solution to a real-world problem.

Wifi-Based Beaconless Human Rescue in Avalanche Disasters

Steven Cho, S.B. ‘24, Electrical Engineering

Advisor: Ninad Jadhav and Tianhong Li

Please give a brief summary of your project.

My thesis project gives users the ability to detect avalanche victims buried under the snow even if they do not have avalanche transceivers on them. It does this by detecting aluminum, a commonly found metal in winter equipment.

How did you come up with this idea for your final project?

After seeing MIT's Katabi Lab and Harvard's REACT Lab do abundant Wifi localization/detection research, I was inspired to identify and address my own WiFi detection problem in the real world.

Is there a real-world challenge that this project addresses?

This project looks to address the demographic of avalanche victims who do not possess avalanche transceivers. There are currently no industry-grade products specialized for transceiver-less avalanche victim rescue.

What was the timeline of your project?

There was the hardware component and software phase for my project. The hardware took about two months to finalize, and the software took about two additional months to finalize.

What part of the project proved the most challenging?

The most difficult component of my thesis project was creating a robust aluminum detection algorithm. The challenge specifically was finding the optimal input-data augmentation and classification machine learning pairing. When designing and building a machine classification algorithm, the most common cause of failure is overfitting too closely to training data. Therefore, for the machine learning algorithm to effectively classify on unseen test data, I had to explore and test countless combinations of data preprocessing and classification algorithms. This took consecutive weeks of testing and reiterating to converge on the final design solution.

What part of the project did you enjoy the most?

I actually enjoyed the struggle to generalize my ML algorithm. I applied a lot of the technical concepts taught in courses, and it is now forever ingrained into my head.

What did you learn, or skills did you gain, through this project?

Overall, I'd say I learned a lot about long-term time management. Knowing how to allocate how many weeks to each stage in the product design process would have streamlined my process much more efficiently, and I learned this the hard way in the last few weeks. In terms of skills I've gained, I now have an intuitive sense of the whole machine learning development pipeline, signal processing, and avalanche safety!

Press Contact

Matt Goisman |