Hardware grants for the parts you need to ship.

A privacy-conscious machine learning system to estimate occupancy and predict wait times in the Silliman Acorn. The system does not use facial recognition and instead counts the number of people present in the space.

The model runs on a Raspberry Pi with an attached camera module. To protect privacy, faces are obfuscated using Gaussian blur, no live video stream is shared, and no image or video data is stored. The system transmits only aggregate information — estimated occupancy and predicted wait time — to a public-facing website.

A ridge regression model trained on collected wait-time data predicts expected delays from observed occupancy, letting students assess crowd levels before arriving. Coordination with Silliman facilities is already in place for deployment.

A low-cost indoor air quality monitor that forecasts pollution levels rather than only reporting them. Users can act before conditions degrade — opening a window, swapping an HVAC filter, stepping out of the room — instead of reacting once the air is already bad.

Predictive modeling of indoor air quality has been validated in research but isn't yet built into consumer devices. This project brings that approach to a hardware form factor priced for everyday use, helping people understand how air pollution factors into their health.

A miniature Baymax-inspired health companion. The device takes voice input describing symptoms and scans a patient's body or face using open-source LLMs and vision-language models, returning observations through a small touchscreen interface.

The software stack is functional; the grant covers the hardware needed to bring the system into a portable, embodied form — a Raspberry Pi unit with camera, microphone, speaker, and 5-inch touchscreen, packaged into a custom enclosure.