30 incredible teams building on Hyperbolic joined us at CalHacks 2024. They each demonstrated how they are leveraging our GPU marketplace and inference services to create applications that tackle real-world problems. From disaster relief to school safety, and even interactive educational toys, each team showcased their own unique and impactful solutions that embodied our vision of a collective AI ecosystem building a better world.
Hyperbolic’s goals for CalHacks were simple: support developers in realizing their visions, identify challenges they faced using our platform, and learn how we can continue to improve our services to better support real-time, large-scale AI applications.
Out of an impressive lineup of winning projects leveraging Hyperbolic, three stood out to us not only for their technical achievement, but also for their powerful missions to make a difference. Here’s a closer look at these winning teams and what they built:
1. CLaiM – Disaster Recovery Made Easy
After witnessing the devastating effects of hurricanes on one team member’s home in Florida, the CLaiM team set out to simplify the insurance claim process. For disaster victims already emotionally overwhelmed by the damage to their homes, proving the existence and value of lost items can be extra challenging. CLaiM eases this process with a comprehensive item cataloging and valuation app.
CLaiM allows homeowners to record a video of their house, and using Hyperbolic’s AI inference for item recognition and value estimation, the app catalogs all household items and stores their details on the blockchain for verification. After a disaster users can rescan their homes and CLaiM will automatically suggest items to include in their insurance claim based on before-and-after comparisons.
The app integrates YOLOv8 and Meta’s Segment Anything Model (SAM) for item segmentation and utilizes Hyperbolic’s Qwen2 VL7B inference model to power the valuation component while ChromaDB stores vectorized item data. These technologies, brought together on Hyperbolic, provide robust item tracking capabilities to provide disaster victims with the support they need.
2. Watchdog – Real-Time School Safety
Tragic statistics around youth gun violence and school shootings drove the Watchdog team to design an AI-based security solution to proactively protect students and staff.
Watchdog uses AI to detect weapons and monitor emotional escalations from live streams of school surveillance footage through audio sentiment analysis. When the system identifies signs of distress or a potential threat, it sends instant alerts to security personnel for quick intervention.
Watchdog uses Ultralytics YOLOv8 trained on the Roboflow platform for weapon detection and Hyperbolic’s Inference API to power real-time data processing for weapon and sentiment recognition. Firebase and Convex support rapid data retrieval, and Roboflow.js runs on the client for low-latency detection in-browser.
Watchdog creates safer learning environments by providing a proactive safety net for schools through enhanced threat detection for faster response times to potential incidents.
3. Teddy AI – Educational and Interactive Toys
The Teddy AI team was inspired by watching their young family members sadly lose interest in their toys. Remembering how much joy their own toys brought them as kids, the team wanted to develop a new kind of interactive toy that would keep kids engaged while adapting to children’s evolving educational needs.
Teddy AI brings toys to life by embedding a Raspberry Pi, microphone, and speaker in a stuffed animal to create an interface where children can talk to their toys. Leveraging prompt engineering and the Hyperbolic Inference API, Teddy AI can lead kids through math lessons in a “choose your own adventure” format, making learning a fun and interactive experience.
Teddy AI uses Hyperbolic’s Audio Generation API and the Hume API for conversational capabilities as well as Hyperbolic for generating interactive story images and audio. The team engineered a prompt-based learning interface that tracks educational progress, providing personalized interactions based on each child’s learning pace.
By blending entertainment with education, Teddy AI turns toys into teachers, supporting childhood development in a unique and engaging way.
The Hyperbolic Community Builds Better Together
We’re incredibly proud of every team that built on Hyperbolic at CalHacks 2024 and cannot wait to see where their projects go from here. As we look ahead, we’re committed to refining our platform to better support real-time AI applications.
We’re also excited to be supporting more real-world solutions building on Hyperbolic at Open AI Day and ETHBangkok in November.
Thank you to everyone at CalHacks 2024 for making this event unforgettable and leveraging Hyperbolic to power your ideas!
