Navi AI raises over $6m and unveils AI for pilot training

Navi AI has formally launched the commercial roll-out of its generative AI platform for training flights. The San Francisco-based start-up, founded in 2024, says its system has been trained on more than 100,000 hours of real-world flying, and will begin deploying this spring across leading aviation training centres. The aim is to automate a full post-flight debrief after every sortie — not just following incidents — to speed up learning and standardise performance assessments.

During a flight, the platform ingests cockpit audio, telemetry and a range of operational and environmental sources, from weather and aircraft histories to surrounding traffic. A domain-specific large language model analyses intent, behaviour and performance before producing an interactive report with 40 to 50 key takeaways. It breaks the lesson down phase by phase, from engine start to shutdown. Instructors and trainees get a consistent, actionable view of gaps and good practice, and can pinpoint areas for improvement. The company has also struck a technology partnership with Garmin for direct integration with its avionics.

"Aviation safety has improved dramatically over the decades, but has for the most part been reactive: We wait for things to go wrong to look at the data and understand why," said Nikola Kostic, co-founder and chief executive of Navi AI. "With Navi AI, every maneuver, every callout, every training flight becomes data that teaches how to make the next one safer and more efficient."

To build out its platform, Navi AI has raised more than $6 million (around €5.17 million): $1.5 million in pre-seed funding in April 2024, $3.35 million in seed funding in March 2026 led in particular by United Airlines Ventures, and a $1.27 million SBIR grant from the U.S. Department of War to adapt the platform for the U.S. Air Force Test Pilot School at Edwards Air Force Base. The technology is already in use or under evaluation at Sling Pilot Academy, the University of North Dakota, Purdue University and Delta State University. The start-up employs seven people and plans to hire a further 15 in 2026, before taking the model into commercial aviation.