Amelia

Inspiration

Air traffic control is chronically understaffed and results in many delays, safety violations, and fatal accidents every year. This has been exacerbated by the current government shutdown during which over 40% of ATCs have stopped working (and up to 80% at some locations).

This is disastrous.

It can take up to 7 years to train an air traffic controller, so bolstering the workforce would take a concerted multi-year effort. Yet, with high churn from existing ATCs, it's likely that the workforce would reach a critical level by then.

But what if we could make every controller 10x more effective?

Not by replacing them—the human in the loop is essential for a task this critical—but by augmenting them. We built Amelia as a second brain for ATCs: surfacing missed insights from real-time flight and weather data, automating routine communication, and reducing the cognitive load that leads to fatigue and errors. Like Cursor for air traffic controllers, we're not removing the ATC—we're making them superhuman.

We offer Amelia as a second brain for ATCs, turning them from 1x to 10x ATCs by surfacing missed insights from real-time flight and weather data and automating controller-pilot communication. We recognize that it is essential to have a human in the loop, so we've focused on giving ATCs superpowers, not replacing them.

What it does

Amelia aggregates data from various real-time airline, weather, and predictive data sources and identifies possible risks for all flights within a 40 nautical mile radius of a target airport. Amelia autonomously surfaces these insights directly to its air traffic controller and automates the voice communication with relevant pilots.

How we built it

We scraped NTSA accident/incident reports from the last two years to identify the most common factors that lead to accidents/risks. Afterwards, we studied air traffic controller procedures using the NTSA ATC manual.

Then, we built Amelia.

On the backend, it uses Grok, Dedalus Labs, and ElevenLabs agents to identify risks, generate insights according to standardized FAA phraseology, and swiftly alert the pilots.

On the frontend, we used React, Typescript, and Tailwind CSS.