Guardian Angel

Inspiration

We have a pilot in our group who has had a near miss or two. We were discussing ideas and noticed all the planes taking off and landing at Logan airport and the idea popped into our head. Especially after having conversed with some folks from the west coast who were talking about eVTOLs, delivery drones, etc.

What it does

It ingests structured and unstructured data, converts them both into actionable data, then evaluates them against a "compliant" state. Death/a high risk of injury is non-compliant. If at any point the data indicates the plane is likely to become non-compliant (by disobeying ATCs commands, not paying attention, or just making a simple mistake).

How we built it

A lot of energy drinks. We use transformer models running on the jetson to turn unstructured data into actionable data. That was one of the key challenges. We don't have too much time to answer these, but we've uploaded the repo and our powerpoint which contain more details.

What's next for Guardian Angel

Optimization. The platform is capable of being optimized to the point where it can run on newer phones and other edge devices, massively increasing adoption.

Built With

• cpp
• cuda
• javascript
• jetson
• lllama
• python
• pytorch