AI (Artificial Intelligence) Training Platforms
Artificial intelligence is fast becoming an essential tool across many industries, and the WorldTour peloton is no exception.
AI is increasingly making its way into cycling. More and more riders within the Talent Performance Group are using digital AI trainers. Although the ‘real’ coaches are still in the majority, this development continues steadily. For young riders who cannot afford an expensive personal trainer, this offers a valuable solution. Thanks to AI, the Talent Performance Group can also provide more talented riders with free performance coaching and management.
Top AI Training Platforms
- Vekta – For pro teams or those working closely with a coach; combines AI with human guidance. Used by teams like Jayco–AlUla, FDJ Suez, Arkéa–B&B Hotels, and GreenEDGE Cycling.
- TrainerRoad – For structured, scientifically grounded progression.
- Xert – For in-depth, data-driven, adaptive training.
- Spoked – For riders with changing schedules needing daily flexibility.
- JOIN (TrainingPeaks) – For planning, tracking, and coach-led training plans.
From predictive nutritional analysis to mathematical modeling, teams are turning to futuristic tech to streamline processes and deepen analysis, ultimately aiming to increase performance.
AI in cycling apps is not new. TrainerRoad has been using machine learning since 2021 to detect a rider's FTP. Xert and others like Spoked and JOIN have also established themselves in recent years.
Google applies dogfooding practices and a Beta release approach for their Office tools. Canary releases are used internally before global deployment. Extensive internal testing is performed by Google test teams.
Methodology:
- Dogfooding
- Continuous delivery
- Almost everything developed on mainline
Environments:
- Small web apps: Laptop > STG > PRD
- Office tools: DEV > TST > Canary > Beta > PRD
Release frequency: Many services see releases multiple times a week
Dogfooding at Google
Google Release Pipeline Video
Release frequency: Production releases twice a day
Amazon
Environments: DEV > STG > PRD (for digital music retail website)
Release frequency: Changes to production every 11.6 seconds on average (May 2011)
Spotify
Workflow: Local DEV > STG > Canary > PRD.
Manual deployments involve DEV accompanying releases to OPS in person for rapid “get it live” delivery. Containers are widely used in production with Docker images built, tested, and reused in production.
Catfooding: Using competitors’ products
Spotify Software Engineering Careers
Unbounce – QA and DevOps
Unbounce integrates QA and DevOps closely: QA owns the responsibility to push code to production. Blue-green deployments are push-button easy. QA verifies, cuts over, and can roll back if needed.
Key principles:
- QA owns continuous improvement and quality tracking across development
- Tests are code; automation is essential
- Testers are quality advocates, improving repeatability and predictability
- Functional, load, stress, and performance testing are integrated
This page illustrates how AI training in cycling and DevOps practices at major tech companies use continuous delivery, QA integration, and predictive analytics to accelerate performance and reliability.
