Executive Summary
Peritos Solutions designed and deployed a cloud-native, AI-powered mobile application for Yorker Limited, an Auckland-based sports-tech company dedicated to improving cricket bowler performance and safety.
Built entirely on AWS, the app uses AWS Bedrock foundation models to deliver personalised training suggestions, injury-prevention insights and natural language Q&A for players and coaches. It fills a critical gap in cricket training by replacing guesswork with real-time, evidence-based recommendations for tracking bowling loads and preventing overuse injuries.
Project outcomes
Results & Impact
AWS Bedrock
Foundation Model Engine
Real-Time
AI Training Suggestions
0
Third-Party AI Costs
100%
Serverless Architecture
Background
Grassroots, club and academy bowlers rarely have access to the workload management tools used by professional teams. Without structured tracking, they often over-bowl in the nets, leading to stress fractures, shoulder injuries and burnout, while coaches rely on manual scorebooks and memory.
Yorker was created to democratise performance science for cricket. Peritos was engaged to build the full technology stack: Android and iOS apps, a serverless AWS backend, and an AI layer powered by AWS Bedrock that answers training questions, suggests improvement pathways and flags injury risks.

Challenge
Early Bedrock responses lacked cricket-specific terminology and context, and response times were too slow for a smooth user experience.
Calculating acute-to-chronic workload ratios was complex because players logged sessions irregularly and data was often missing. Many players also practise in areas with limited connectivity.
Every player needed different AI context, so personalising responses at scale was a key challenge.
Solution
The platform runs on a 100% serverless AWS architecture in the Sydney region, with four layers: the mobile app, API and compute, AI intelligence, and data and notifications. API Gateway and Lambda handle requests, Cognito manages authentication, DynamoDB and S3 store player and training data, SNS delivers push notifications and CloudWatch provides observability.
Lambda functions call Bedrock with structured player data and custom-engineered prompts. Peritos built a prompt library covering speed improvement, load management, injury prevention and performance analytics, with each template pulling player context from DynamoDB.
Key Solution Capabilities
- Bowling Load Tracker: Log sessions by delivery type, intensity and volume, with acute-to-chronic load tracking.
- AI Speed Plans: Receive personalised four-week plans with drills, strength exercises and technique cues.
- Injury Risk Alerts: Get green, amber or red risk ratings with explanations and rest plans.
- Natural Language Q&A: Ask questions like “Is my workload safe this week?” and get answers based on personal data.
- Weekly Insights: Receive AI-generated performance summaries compared with historical averages.
- Coach Dashboard: View squad-wide loads, AI risk flags and downloadable session reports.
- Smart Reminders: Stay engaged with SNS nudges to log sessions, rest or review plans.
Approach
Peritos followed a 12-week, sprint-based delivery model. It moved from requirements workshops and architecture design through authentication, load tracking, Bedrock integration, the coach dashboard and notifications, to end-to-end testing, UAT, go-live and hypercare.
To solve the challenges, Peritos added domain-specific context to prompts and built Lambda rolling-window calculations with data imputation. It enabled local session caching with DynamoDB sync once connectivity returns, and reduced latency through Lambda optimisation, connection reuse and streaming responses. A DynamoDB-backed player profile injects each player’s history, style and goals into every prompt.
As a result, grassroots bowlers gain access to elite-level guidance, and coaches save time on manual tracking. Yorker scales automatically at minimal cost, with no third-party AI subscriptions and no player data shared with third parties.









