Data Scientist - Aviation

• PhD in machine learning, statistics, aerospace engineering, operations research, or a related field, or an MS with at least 3 years building models against real operational data \n

• Aviation industry background with direct work experience involving the FAA, NASA, airports, airlines, or air traffic management. This can come from ATC, flight operations, airport operations, avionics, or prior aviation research, but it needs to be deep enough that you can tell when a model output makes no operational sense \n

• Strong probabilistic modeling, including Bayesian inference, state-space and sequence models, and uncertainty quantification. We weight calibration more heavily than benchmark accuracy \n

• Demonstrated depth in at least one of the following: time-series or trajectory prediction, speech recognition and NLP, computer vision and multi-object tracking, or sensor fusion and target tracking \n

• Fluency in Python and the modern ML stack (PyTorch, scikit-learn, pandas), with experience on large and imperfect operational datasets, * A disciplined approach to model failure, meaning you characterize how a model behaves when it is wrong and report it early

We are looking for a Data Scientist to lead modeling on our NASA SBIR research in in-time aviation safety. The work turns live surveillance tracks, SWIM feeds, ATC audio, weather, and NOTAMs into systems that predict unsafe operations on airport surfaces and in terminal airspace before they become incidents. The problems are real-time, multimodal, and uncertain, and none of them have settled solutions.

\n

You will own the models: state estimation, event prediction, and the speech and vision stack. This is a research role with operational consequences, so we need someone who understands both the math and the airport.

\n

\n

What You Will Do

\n \n

• Develop probabilistic models (HMM and GMM class) that infer aircraft activity from noisy, sparse, and irregularly sampled position data, and that report calibrated confidence as surveillance quality degrades \n

• Forecast when an aircraft will reach a runway, cross a hold-short line, or begin its takeoff roll, with quantified uncertainty. Alert lead time is the core value of the system, and a prediction without a defensible error bound is not operationally usable \n

• Fine-tune ASR models (Whisper with LoRA) on live ATC audio and build the parsing layer that converts a transmission into structured intent (callsign, taxi route, hold-short, crossing authorization) at roughly one second of latency \n

• Associate tracks across ADS-B, computer vision detections, and parsed voice clearances that arrive at different update rates and on different clocks (JPDA-class association, EKF state estimation), so that deviation from a cleared taxi route is detected as it happens \n

• Detect and track aircraft, vehicles, and ground equipment that broadcast no position at all (YOLO-class detection with ByteTrack), closing the blind spot left by cooperative surveillance \n

• Build and defend the performance evidence presented to NASA, including alert lead time, false alert rate, degradation behavior, and model behavior when an input is lost \n

• Contribute to technical reports, program briefings, publications, and follow-on proposals \n

\n

\n

Cignus Consulting is an SBA 8(a)-certified, MBE/DBE aviation technology and engineering firm headquartered in Leesburg, Virginia. Since 2007 we have supported the FAA, NASA, DoD, airport authorities, and aviation clients with systems engineering, airspace and airside simulation, aviation data analytics, and software products built on standards like SWIM, ADS-B, and ARINC 424. Our work ranges from federal task orders and SBIR research to commercial simulation and digital twin platforms.