Case study

Nudge Pet

A closed-loop biofeedback companion that reads your stress from a wearable sensor and answers through a virtual emotional support animal, built in 2.5 days and named a Top 10 finalist at MIT Reality Hack 2025.

Type
Hackathon · Mixed-reality biofeedback
Role
Product Manager + Research Lead
Tools
Unity · C# · Blender · EmotiBit · Meta Quest 3
Timeline
2.5 days · MIT Reality Hack 2025
Team
5 people · PM, 3D, XR, hardware
BiofeedbackResearchMixed realityMental health
Design
Divergent → convergent ideation → a closed-loop biofeedback concept
Research
Animal-assisted therapy + wearable biosensing (EDA / PPG / temperature)
Build
EmotiBit ↔ Unity integration · live user testing on the floor
Problem

Not everyone can have an emotional support animal

Emotional support animals deliver real, measurable stress relief, but ownership is gated by resources, living situation, and physical ability. The people who might benefit most are often the least able to keep an animal.

This one started personal. Nudge Pet is modeled on Snuffles, my own emotional support animal, and the question we kept circling was simple: what if caring for yourself felt as easy and joyful as caring for a pet?

2.5days from idea to a working biofeedback demo
Top 10finalist of hundreds of teams at MIT Reality Hack 2025
3physiological signals fused into a live stress estimate: EDA, PPG, temperature

Our research question: could a virtual companion deliver some of that judgment-free comfort, and could we make it respond to a person's actual physiology in real time, not a button press?

Snuffles, Emmanuel's real emotional support animal
The real Snuffles. The whole concept is an attempt to give other people a sliver of what an emotional support animal provides, without the resources or living situation ownership demands.
Approach

A closed loop between body and pet

Nudge Pet is a biofeedback loop you can feel. An EmotiBit wearable streams electrodermal activity (EDA), heart rate via photoplethysmography (PPG), and skin temperature into Unity, where those signals become a live estimate of the user's stress. The virtual pet, a passthrough-MR companion on Meta Quest 3, responds to that estimate: calm when you're calm, and when stress spikes, it gently pulls you into a guided meditation.

Two design choices did the heavy lifting:

Why a pet. The research on human–animal interaction is consistent: animals are perceived as non-judgmental, which lowers the barrier to confiding and self-soothing. A pet is an emotional interface people already know how to use. No instructions required.

Why mixed reality. Passthrough MR keeps the user grounded in their real room rather than sealing them into a headset. For a wellness tool meant to reduce anxiety, presence beats immersion. The pet shares your space instead of replacing it.

Jessica Sheng modeling and rigging Snuffles in Blender
Jessica Sheng rigging the virtual Snuffles in Blender. The model was custom-built and driven by a state machine, so the pet's behavior could track the user's stress state.
EmotiBit biosensor signal breakdown: EDA, PPG, and temperature
The EmotiBit signal stack. EDA is the most effective real-time anxiety measure; PPG adds heart rate and respiration. Reading these live was the core technical bet of the project.
Process & artifacts

From divergent ideation to a working loop in 2.5 days

Ideation. We opened wide, a divergent session exploring every way biosensing could enhance emotional wellbeing in XR, then ran a convergent pass to commit. The winning idea had to clear two bars: genuinely useful, and buildable by five people, none of whom had touched hardware before, in under three days.

Divergent ideation session: wide range of biosensing concepts
Divergent: every direction biosensing could take an emotional-wellbeing experience.
Convergent ideation session: narrowing to the virtual pet concept
Convergent: narrowing to the virtual emotional support animal, scored on impact against what we could actually ship in 2.5 days.

Research foundation. I grounded the concept in three findings: animals are perceived as non-judgmental, which lowers the barrier to self-soothing; computer-simulated pets have been shown to promote empathy and healthier behaviors; and among physiological signals, EDA is the most effective real-time anxiety measure, with PPG supplying heart rate and respiration. Those findings set the sensor priorities and the interaction model.

Integration & testing. The hard part was making the EmotiBit talk to Unity reliably enough to drive behavior live. I helped integrate and test that sensor pipeline against the VR experience, and ran user testing on the floor to confirm the pet's responses actually read as responsive, that people felt seen by it.

Real-time EmotiBit biosensor readings streaming into the experience
Live biosensor readings streaming in, the signal that drives the entire loop.
Mixed-reality guided meditation scene triggered at high stress
Meditation mode, triggered when detected stress runs high: a guided mixed-reality moment rather than a notification telling you to calm down.
Real-time stress detectionEmotiBit monitors EDA, PPG, and temperature to estimate emotional state continuously, with no user input required.
Dynamic pet responsesThe companion adjusts its behavior to the detected stress level, the core closed-loop interaction.
Interactive pet careUsers reach out and pet Snuffles; sparkle particle effects answer the touch, rewarding the act of caring.
Immersive meditation modeA guided mixed-reality meditation activates when high stress is detected, meeting the user in the moment.
Polished 3D animationA custom-rigged Snuffles driven by a state machine, so behavior maps cleanly to the user's physiological state.
Impact

Top 10 at MIT Reality Hack 2025

Top 10 finalistSelected from hundreds of teams at MIT Reality Hack 2025, a global hardware-and-XR hackathon
Hardware, first tryEveryone on the team was working with hardware for the first time, yet we shipped a hardware-driven project into the finals
Closed loop in 2.5 daysA full biosensor → stress-estimate → responsive-pet → meditation loop, integrated and demoable inside the time limit
Validated on the floorUser testing during the build confirmed the pet read as genuinely responsive, not scripted
My role

I served as Product Manager and Research Lead. I set the product vision, defined scope, and steered strategy under a brutal time limit, and I led the team through the divergent-to-convergent ideation that landed us on Nudge Pet. As research lead, I grounded the concept in the literature on animal-assisted therapy and wearable biosensing, which set our sensor priorities (EDA first) and the non-judgmental-companion interaction model.

On the build, I helped integrate and test the EmotiBit biosensing pipeline against the VR experience, and I ran user testing to validate and refine the loop. The engineering and 3D work was a true team effort.

Built with Jessica Sheng (3D modeler + project manager), Steven Le (XR developer), Ethan Johnson (EmotiBit developer), and Audrey Lane (project manager). Demo video by Cole Lee. Snuffles is my real emotional support animal.

Reflections & takeaways

None of us had built with hardware before, and the honest lesson is how much of a hackathon is risk triage. We picked the idea we could actually de-risk in 2.5 days, not the flashiest one. The closed-loop bet, driving a virtual pet off live EDA, was the riskiest piece, so we built that spine first and let everything else hang off a working signal. The softer takeaway: anchoring the project in a real emotional support animal, Snuffles, gave the team a true north for every design call. A wellness concept earns its credibility from honesty, not polish, and having a real referent kept us from designing a gimmick.

Nudge Pet. MIT Reality Hack 2025, Top 10 Finalist. Team: Emmanuel Corona, Jessica Sheng, Steven Le, Ethan Johnson, Audrey Lane. Demo video by Cole Lee. Devpost: devpost.com/software/nudgepet · Code: github.com/StevenLe2012/nudge_pet · Demo: youtube.com/watch?v=v8tnNr-b73s