MOCHI — A Little Robot That Cares

Inspiration

Getting help when you don't feel well should be simple, comforting, and accessible. But when someone has a headache, feels under the weather, or isn't sure what to do next, figuring out which supplies to use or whether to contact a healthcare professional can be overwhelming.

We wanted to build something that makes that experience feel more natural. Inspired by the idea of combining an approachable companion with useful healthcare technology, we created Mochi: a voice-enabled AI healthcare companion that brings together conversational assistance, personalized recommendations, and physical interaction.

Instead of being just another chatbot on a screen, Mochi has a physical presence. Users can speak to it, receive spoken responses, and be guided toward the appropriate supplies in its drawers. Our vision is to make healthcare support feel more intuitive, interactive, and human-centered.

What It Does

MOCHI combines an AI-powered app with a physical robot prototype.

Voice conversations: Users can speak naturally to Mochi through a microphone, and Mochi responds aloud through a speaker.
AI health guidance: Mochi is designed to help users understand common symptoms, offer general health information, and recognize when professional medical care may be appropriate.
Doctor recommendations: When a situation calls for professional help, Mochi aims to help users find relevant healthcare providers.
Smart drawer organization: Mochi's three-drawer design organizes supplies into categories, such as temperature-checking tools, comfort supplies, and medications.
Physical interaction: Our design connects the app to the physical robot, allowing it to indicate or eventually open the relevant drawer.
Inventory memory: The app is designed to maintain an inventory of drawer contents so that recommendations can be matched to available supplies.

For example, if a user says, "I have a headache," Mochi could ask follow-up questions, offer appropriate general guidance, suggest checking relevant supplies, and recommend seeking medical advice when warranted. It could then identify the drawer containing the relevant item.

MOCHI is intended to support health-related decisions, not replace a doctor or provide a medical diagnosis.

How We Built It

We approached MOCHI as a combination of software and hardware.

1. The App and AI Experience

The app is the intelligence behind MOCHI. It is designed to interpret user conversations, provide relevant responses, maintain information about available supplies, and coordinate the robot's actions. The goal is to make the interaction feel like talking to a helpful companion rather than navigating a complicated medical interface.

2. The Hardware Prototype

We chose a Raspberry Pi as the controller for the physical prototype. Our available components include a Grove expansion shield, an LCD screen, a breadboard, jumper wires, a webcam, and a speaker. We are using these components to explore a no-solder prototype that can hear users through the computer-connected microphone, respond through the speaker, and communicate with physical indicators and drawers.

The laptop handles the more demanding tasks, including voice processing and AI interaction, while the Arduino is intended to handle physical outputs and drawer mechanisms.

3. The Three-Drawer System

We designed MOCHI around three separate compartments to keep supplies organized. The initial concept includes a thermometer drawer, a comfort-supplies drawer, and a medication drawer. The app's inventory is intended to connect each item to its physical location, allowing MOCHI to guide users to the right compartment.

Our initial approach prioritizes a simple, manually accessible drawer structure, with electronic drawer movement and status detection added as the hardware is validated.

Challenges We Faced

One of our biggest challenges was connecting software intelligence to physical hardware. Building a conversational app is one problem; making it interact reliably with a real robot introduced additional constraints, including access to compatible components, power requirements, wiring, motor control, and communication between the laptop and microcontroller, all within the constraints of the hackathon.

Another challenge was designing the healthcare experience responsibly. A system that responds to symptoms must do more than recommend an item. It needs to recognize uncertainty, avoid presenting guesses as diagnoses, and help users understand when professional medical care is necessary.

What We Learned

MOCHI has taught us to think about AI as a complete interactive system rather than just a model that generates text. A useful experience depends on how users communicate with it, how it remembers relevant information, how software interacts with physical components, and how reliably the system responds.

Most importantly, we learned that building a healthcare companion requires balancing convenience with responsibility. Technology can make health information and everyday supplies easier to access, but it should also recognize its limitations and encourage professional care when needed.

What's Next

Our next steps are to complete the voice interaction loop, connect the app to the Arduino over USB, verify the LCD and sensor wiring, and implement reliable drawer indicators or motorized movement. We also want to build a persistent inventory system, improve doctor recommendations, and refine MOCHI's interface into a polished, friendly companion.

Our vision is simple: healthcare support that feels less intimidating, more accessible, and a little more caring.