Qualcomm RB3 Gen 2 development kit at MIT Reality Hack

Who doesn’t like a good hackathon? Especially when the theme is experiential technology, like Augmented/Virtual/Extended reality (AR/VR/XR)? With categories like “Standing on the Shoulders of Sustainability,” “Pioneering a Neuroadaptive Future” and “Jaw-Dropping Hardware?”

As a sponsor of the 2025 MIT Reality Hack in Cambridge, Massachusetts, Qualcomm Technologies offered hackathon participants the opportunity to build projects around the Qualcomm RB3 Gen 2 development kit. Our Developer Marketing team prepared participants by conducting a pre-hack online workshop, then two in-person workshops at the event for about 60 developers on 12 teams.

Best of all, we awarded $1500 cash prizes in three categories that featured our technology: Best Use of Artificial Intelligence (AI), Best Social Impact and Best in Internet of Things (IoT). We also awarded an internship opportunity on our Developer Marketing team as a tech evangelist.

Hardware and a pipeline of software tools

The Qualcomm RB3 Gen 2 development kit is designed for applications like ruggedized handhelds, industrial scanners, dash cameras and human-machine interface systems. Besides its powerful CPU and GPU, it’s made for edge AI processing and computer vision, with pre-integrated sensors and drivers for camera and motor control.

We gave participants the documentation for the Qualcomm RB3 Gen 2 kit, then we introduced them to a pipeline of software tools that complement the kit:

• Qualcomm AI Hub for AI models that work on the kit
• Sample applications for the kit and the Qualcomm Intelligent Multimedia (IM) Product SDK
• Qualcomm Intelligent Robotics Product (IRP) SDK
• Visual Studio Code IDE

Once teams got the sample apps running and saw the on-device AI inference using the dev kit, they were blown away. The first team to execute the VS Code IDE and see the sample app running was particularly impressed.

And the winner is . . .

The hackathon ran from Friday morning through mid-afternoon Sunday (January 24-26). Of course, we in Developer Marketing worked almost as long and hard as the developers themselves: answering questions, taking notes, suggesting workarounds and birddogging technical issues with our colleagues.

It’s amazing how much progress a team of determined developers can make, coming up with ways to use our software and hardware that had never occurred to us. Here are the winners in the categories sponsored by Qualcomm:

Best use of AI

Neuroveil: The Twin Mind Interface - Synchronize brainwaves to treat mental disease and generate cymatic art.

Neuroveil synchronizes brainwaves between participants, allowing them to converge to a shared brain state and control that brain state to elicit certain effects. The team used data collected from OpenBCI sensors to detect the brainwaves, then trained a custom model on the sensor data to identify the subject from whom the signals originated. Using Python packages like NumPy and Pandas, they successfully ran inference on the Qualcomm RB3 Gen 2 within a Docker container.

Best social impact

FarSight - Bring clarity to the chaos

In the wake of devastating wildfires in Los Angeles, this team created an end-to-end demo of a rover designed to help firefighters. The rover runs an object detection model capable of identifying people, pets and fire hotspots and uses sensors for air quality and temperature to monitor environmental conditions. An Arduino board drives the motors. FarSight runs AI inference on the Qualcomm RB3 Gen 2, sends data to a Redis server and displays it in a VR headset. The team 3D-printed the rover, with mounts for the dev kit and the Arduino board. This was the most complete demo we saw at the event!

Best in IoT

RealityBridge – Quickly prototype VR scenes using AI object detection

This project, a handheld copy-and-paste tool, was inspired by the desire to quickly prototype scenes in virtual reality and lay out models in a virtual 3D space. Users scan real-world objects using the Qualcomm RB3 Gen 2 dev kit like a handheld camera. The kit detects and classifies the scanned object, then publishes the latest-detected, high-confidence object to a Redis server. A VR Unity project subscribed to the Redis server takes the AI-identified name and uses the SketchFab API to spawn a 3D model in virtual reality.

Honorable mention

AutoChef – Eliminate waste by monitoring food items in the home

AutoChef uses AI to monitor food in your refrigerator. The app determines and tracks shelf life, then notifies you before food spoils. It runs on the YOLOv8 model with COCO8 datasets for detection of fresh and rotten produce. Using Qualcomm® Linux® 1.3 software stack with the Qualcomm IRP SDK build, the team took advantage of the qrb-ros-camera pipeline and ROS to view camera frames and sensor data in a VR headset. Their integration shows the potential of combining IoT and VR in industrial use cases.

Qualcomm Technologies’ solutions in other tracks:

MIRAGÉ

TrueForm

Augmented Fire Management Center

ZenFriend XR

Paw Pulse

Your turn

If you’re feeling as inspired by this creativity as we were, you can purchase your own Qualcomm RB3 Gen 2 development kit and start innovating. The links above will take you to the code repositories of most of the MIT Reality Hack projects. We’ve also published a number of sample applications to get you started.

You’ll find another layer of support in our Developer Discord, including deeper insights and real-time conversations with fellow developers. Look for upcoming live interviews on Discord with some of the winners from MIT Reality Hack 2025.

Our Developer teams may soon be coming to a hackathon or conference near you. Check our Developer Events page to see where we’ll show up next. Bring your cool ideas and burning questions.