Babson Buildathon — ReFrame
Embedded software and mechanical design for the ReFrame buildathon project
Led embedded‑software integration and mechanical packaging for the ReFrame prototype. Responsibilities included sensor integration, real‑time control loops, and ensuring mechanical interfaces met manufacturing constraints. Delivered a working prototype and documentation for rapid iteration.
- Embedded control and interfacing (Python, MCU firmware)
- Mechanical packaging in Autodesk Fusion
- Prototype PCB and sensor calibration
The ReFrame Concept: AI-Powered Memory Support
Inspired by personal experiences with Alzheimer's, our team created ReFrame, an AI-powered biomedical device to address the growing need for compassionate solutions for the 6.7 million Americans living with the disease.
Interactive Memory Frame
ReFrame is a digital picture frame that scrolls through family photos when idle to jog memories. When activated by voice, it engages a custom-trained ChatGPT agent to prompt patients to recount memories and show relevant images. It can also facilitate simple cognitive games like crossword puzzles focused on current events.
Technology & Accessibility
We utilized Scribe for speech-to-text input and Eleven Labs for text-to-speech output, ensuring the agent's response is an easily understandable voice. This design, combined with large on-screen text, ensures accessible communication for hearing or visually impaired elderly patients.
MVP Development & Stack
The Minimum Viable Product (MVP) was constructed using a monitor for the user interface, which was connected to a Raspberry Pi handling the logic. Crucially, all the intelligence—the custom ChatGPT agent functionality—was achieved by making API calls to the OpenAI GPT-4 model.
Tools & Methods
Autodesk Fusion, Python, PCB prototyping, bench validation and user testing.