Navi captures everything a pilot sees and hears and turns it into automated debrief intelligence. You're building the AI that makes sense of it all — the models that listen to cockpit audio, identify maneuvers, attribute instructor vs. student behavior, and generate debrief reports that match what a seasoned CFI would catch.
This is a founding AI/ML role. You'll own the intelligence layer that sits at the core of everything Navi does — from audio diarization and speech recognition to maneuver detection, safety event classification, and automated debrief generation. The data is messy, multimodal, and high-stakes. Cockpit audio is mono with overlapping speakers. Avionics telemetry arrives in dozens of formats. ATC comms bleed into crew conversation. Your job is to make machines understand all of it — and get it right, because the output goes directly to pilots, instructors, and military operators.
Navi is deployed at Embry-Riddle, Purdue, UND, Sling Pilot Academy, and the United States Air Force. Your models are already in production. This isn't research — it's AI that flies.
Build and improve the ML systems that power Navi's automated flight debrief — maneuver detection, performance scoring, safety event identification
Develop and refine audio intelligence pipelines — speaker diarization, speech-to-text, cockpit audio separation, ATC communication extraction
Design the AI reasoning layer that synthesizes avionics data, audio, and ADS-B into coherent sortie narratives
Build evaluation frameworks and feedback loops that continuously improve model accuracy against real-world CFI assessments
Work directly with pilots and instructors to ground-truth model outputs and close the gap between what the AI sees and what actually happened in the aircraft
Push the boundary on what's possible with LLMs in safety-critical, domain-specific applications
5+ years of professional experience in machine learning, AI, or applied research — with production systems, not just papers
Deep expertise in at least one of: NLP/speech processing, audio ML, time-series analysis, or multimodal reasoning
Experience building and deploying ML pipelines end to end — data ingestion, model training, evaluation, inference, and monitoring in production
Strong engineering fundamentals — you can build the infrastructure your models need, not just the models
You've worked with messy, real-world data and know how to build systems that are robust to noise, edge cases, and domain drift
Experience with LLMs — fine-tuning, prompt engineering, retrieval-augmented generation, or building LLM-powered applications
You ship. You don't wait for a perfect dataset or a clean abstraction. You build, evaluate, iterate, and improve
Comfortable operating with high autonomy in a fast-moving environment where the problem definition evolves weekly
Familiarity with aviation systems, flight training operations, or defense technology environments
Experience with audio diarization, speaker separation, or cockpit/radio audio processing
Background in safety-critical ML systems where model accuracy has real-world consequences
FOQA sees a perfect steep turn. Navi sees a disaster. The difference is the intelligence layer you're building — the AI that hears the instructor take the controls two seconds before a G-load exceedance, that knows the student was coached through an approach instead of flying it solo, that catches what the numbers alone will never tell you. This is the system the aviation industry has never had. You're building it.
Early-stage equity — real ownership in a category-defining company
Flight training — earn your pilot's license and build with true domain expertise
Impact you can see — your work will be used by pilots, flight schools, airlines, and the U.S. Air Force
A role that scales into technical leadership as we grow
Find a way. We don't wait for permission or perfect information. Ideas come from anywhere regardless of title. Figure it out, ship it, iterate.
Creativity over control. First principles over process. We'd rather have a creative solution that's 80% right today than a perfect one next quarter.
Update fast. Come in with a hypothesis, throw it away when the data says otherwise. Ego has no place here.
Intensity with focus. We work hard because the mission demands it. Clarity on what matters is how we make that sustainable.
