-
-
Starting Page
-
Result of analysis
-
Recommend action
-
Memory Log
Inspiration
We were inspired by real stories where pets were unintentionally harmed because their behavior was misunderstood, as well as by friends who love their pets but cannot be present all the time due to daily responsibilities. Using Gemini 3’s multimodal reasoning, we aim to help owners better understand everyday pet behavior and receive timely, non-medical alerts when something seems unusual. PogAI keeps an eye on your pet when you can’t—helping you stay informed and act with confidence.
What it does
PogAI is a behavioral analysis platform that leverages the Gemini 3 model's advanced spatial-temporal reasoning to decode pet intent from video and audio sequences. By analyzing complex physical cues, vocalizations, and movement patterns in real-time or from uploaded clips, the application identifies specific animals using personalized profiles and translates their behaviors—ranging from subtle signs of anxiety to clear demands for play—into ranked intent hypotheses with specific temporal evidence traces. The system effectively acts as a bridge between species, maintaining a long-term "memory log" to establish behavioral baselines and flag anomalies that deviate from a pet’s normal routine, while offering actionable, AI-driven recommendations to help owners respond more effectively to their pets' emotional and physical needs.
How we built it
We use AI studio to initialize the application. At the core of the application, we use the capabilities of Gemini 3 multimodal to analyze videos and distinguish each pet. Then, with its reasoning capabilities, we use it to logically examine pets behavior and compare it to the usual behavior. With the power of AI studio and Gemini 3 features, we are able to build a demo version of the application using only prompts.
Challenges we ran into
One of the most challenging things we met was dealing with finding different kinds of pet videos to test the application. This is to make sure our application is robust to a wide variety of pets. Another challenge is about controlling AI studio behavior to ensure a precise and correct application is delivered.
Accomplishments that we're proud of
PogAI is capable of analyzing a pet's behavior frame by frame and precisely distinguishing each animal. This capability has generated excitement regarding the potential of Gemini 3 to develop more sophisticated applications. Furthermore, the timeframe allocated for our project was limited, as we became aware of this hackathon at a late stage. Despite this constraint, we successfully completed the project.
What we learned
We learned and were astounded by Gemini potential. AI studio is a great tool to start making your demo application and give us a head start in making a demo. However, to use it to its full power, users should consider best prompt practices and scope the contexts for the application for precise results.
What's next for PogAI - Track Your Pets Behavior
The next step of PogAI is to optimize tokens used by Gemini 3 from analyzing frames to reasoning results. We believe that to sustain our application in the long-run, a well-optimized application is required. We also plan to combine it with IoTs devices to increase accessibility and utility.
Log in or sign up for Devpost to join the conversation.