Pytho AI

What it does

Our application empowers humanitarian assistance and disaster response teams with the capability to rapidly and effectively deliver critical aid where it's needed most. Emergency response teams often rely on tedious manual processes to assess damage, coordinate resources, and communicate vital information. For example, when identifying the hardest hit areas from a wildfire or a hurricane, first responders spend too much time reviewing drone and satellite images to coordinate a response effort.
Our application automates these time-consuming processes to deliver actionable plans to emergency response teams. Leveraging real-time and multi-modal data sources including social media and sensor streaming data, Pytho can plot a safe rescue route for teams to reach victims, recommend areas to prioritize rescue efforts, and optimize resource employment to save critical time and materials.
We use Gemini API for:
1) Building our cloud-based user facing application.
2) To build a workflow manager that can extract statement of work from a video capture. Similar to ÉCLAIR (arXiv:2405.03710v1)
3) To create an instrumented multi-modal training set to instruction tune small LLMs such as Gemma-2-2b. Some of the use cases included in the training sets are:
- Understanding maps and documents to create geo-based products.
- Extracting location from images.
- Assessing infrastructure damage from images and videos
- Generate products and recommendations for first responders