FlightAware, the Houston flight-tracking company owned by Collins Aerospace since 2021, uses machine learning to predict when flights will arrive and depart. JetBlue runs its Foresight predictions across its whole operation.
Founded in 2005 with 201 to 500 staff, it sells Foresight, which trains more than 1,100 models at once to produce estimated arrival times, the Firehose live data feed, the AeroAPI data service and GlobalBeacon, which helps airlines meet ICAO's global tracking rules introduced after MH370.
Frankfurt (FRA) was Foresight's first client, in 2019. Most evidence for Foresight comes from FlightAware's own channels, and claims of use at hundreds of airports are the company's own.
Aviation AI deployments and updates
01 / Deployed · full scale
FlightAware Foresight predictive flight data from Collins Aerospace across JetBlue operations
JetBlue integrated FlightAware Foresight into its system-wide operations from July 2025, Collins Aerospace, the RTX business that sells it, announced on 26 August 2025. Foresight is a predictive analytics tool fed by machine-learning algorithms and billions of data points, and it alerts operators to likely delays caused by weather, congestion or other factors.
JetBlue uses the early warnings to adjust gates, crews and schedules and to keep passengers informed of changes, according to the release. Tom Lloyd, JetBlue's Director of SOC Strategy and Continuous Improvement, is quoted on its accuracy.
Neither company has published on-time performance or prediction-accuracy results from JetBlue's use. The figures in the release describe the product, not JetBlue's results.
Fraport announced on 10 July 2019 that it would start using FlightAware's estimated landing times, or ELDTs, delivered through the Firehose data feed, as an additional data source for running Frankfurt Airport. FlightAware's engineering blog says Frankfurt became the first client of its predictive technology team by mid-2019.
According to the 2019 release, the predictions are based on statistical analysis of the tracks and timestamps of hundreds of thousands of flights; FlightAware now describes them as coming from machine-learning models. The company says Fraport feeds the predictions into its operational tools so ground teams can position resources around expected arrival windows.
That account of current use is FlightAware's, written in the present tense. Fraport has not commented publicly since 2019, and accuracy figures for Frankfurt have not been published.
A predictive analytics platform that uses machine learning and neural networks to generate real-time flight arrival and departure predictions.
The problem it addresses
Inaccurate or late arrival and departure time estimates lead to inefficient gate preparation, missed passenger connections, wasted ground crew allocation, and excess fuel burn from aircraft holding. Foresight provides early and accurate predictions to enable proactive operational decisions across the gate-to-gate flight cycle.
Key features
Estimated Landing Time (ON/ELDT) predictions available hours before arrival
Estimated In-Block Time (IN/EIBT) gate arrival predictions
Taxi-out duration estimates beginning 3 hours before pushback
Arrival runway probability predictions (updated continuously throughout flight)
4 more features
What-If Gate Change simulation tool showing ETA impact of gate reassignment
Neural network model providing truly global airport coverage
Accessible via AeroAPI (Premium tier) and Firehose streaming feed
Foresight Labs exploration interface for evaluating prediction accuracy
Technical specifications (8)
Neural network architecture providing global single-model predictions
Trained on petabyte-scale compressed historical flight tracking data
Over 1,100 training models (covering ON and IN predictions for 554+ airports as of 2020)
Live streaming prediction environment processing over 1.6 million evaluations per cluster per second
Redundant on-premises deployment across two datacenters
AWS cloud-based model training pipeline (Apache Spark, Kubernetes, Argo workflows)
Machine learning model retraining cadence to address model drift
High-availability SLA with cross-datacenter redundancy
An enterprise-grade streaming data feed delivering real-time, historical, and predictive global flight tracking data.
The problem it addresses
Enterprise aviation stakeholders, including airlines, airports, air traffic management tool providers, and third-party application developers, require a non-stop, low-latency stream of comprehensive global flight data for real-time operational systems. Firehose eliminates the need to query multiple disparate data sources by providing a single, continuously updated, customizable feed.
Key features
Non-stop streaming of real-time global flight tracking and status data
Historical flight data playback capability
Optional inclusion of FlightAware Foresight predictive ETAs and taxi-out estimates
Airport surface movement data (AirportAware) for hundreds of airports
5 more features
Unified feed of airborne and surface movement data (launched November 2021)
Ready to Taxi surface movement data
Customizable data layer selection (radar, ADS-B, ACARS, satellite, Foresight)
Firestarter open-source Docker-based development accelerator for integration evaluation
Best-in-class SLA with 100% service availability recorded for H1 2020
Technical specifications (9)
Powered by HyperFeed, processes over 80,000 aircraft position messages per second
A REST-based on-demand flight data API providing access to real-time, historical, and predictive flight information. Offers over 60 distinct endpoints covering flight status, aircraft positions, tracks, airport activity, operator fleet data, and predictive ETAs via FlightAware Foresight (Premium tier).
The problem it addresses
Organizations that need flexible, on-demand access to aviation data without the continuous stream volume of Firehose require a query-based API with pay-per-use economics. AeroAPI provides structured, developer-friendly access to the same underlying data with full historical depth and configurable scope.
Key features
Over 60 distinct REST API endpoints
Flight status and real-time tracking by tail number, origin/destination, aircraft type, altitude, or groundspeed
Historical flight data from January 1, 2011 onward
A GADSS-compliant (Global Aeronautical Distress and Safety System) global flight tracking and alerting platform for airlines and aircraft operators.
The problem it addresses
Following the disappearance of MH370, ICAO mandated through GADSS that airlines maintain real-time global tracking of all aircraft with position updates at defined intervals. Traditional terrestrial ADS-B and radar systems cannot provide coverage over oceans and remote areas. GlobalBeacon addresses this gap by using space-based ADS-B to provide truly continuous global coverage.
Key features
100% global coverage via Aireon space-based ADS-B on Iridium NEXT constellation
Minimum 1-minute position update intervals, exceeds ICAO GADSS SARPs
No costly avionics modifications required for ADS-B-equipped aircraft
Real-time situational awareness and distress alerting
3 more features
Quick fleet-wide implementation
Monitoring and alerting interface
GADSS compliance certification support
Technical specifications (6)
Space-based ADS-B data source: Aireon hosted on Iridium NEXT satellite constellation (66 low-Earth orbit satellites)
Position update rate: minimum once per minute globally
Aircraft-to-system latency consistent with space-based ADS-B relay
Surveillance grade: exceeds GADSS SARPs for all ADS-B aircraft
Web-based mapping interface
No hardware installation required on aircraft
Certifications and approvals
ICAO GADSS compliant, exceeds Global Aeronautical Distress and Safety System Standards and Recommended Practices
Profile data from the Aerostack companies register; deployment records from the verified register. Evidence grades: T1 first-party confirmation, T2 credible reporting. See Our approach for how records are graded.
Aerostack companies register
Aerostack tracks AI deployments across airports and airlines, with the published evidence behind each record. Photography is real aviation imagery. Record thumbnails may show the named airport or airline; other photographs do not.
