Case Study for Cricket Training app using AWS as a backend for Machineroad

Machineroad Case Study

In this case study, we helped machineroad by developing a bespoke mobile-based application to help improve the Cricket bowling skills for their user base. The app helps to measure the ball speed it allows user to do a video recording and creates an image snippet for the end user showing trajectory.  This was deployed using AWS cloud for storing all backend and front end code. 

We also added detailed level analytics to help them see their progress week on week and month on month and some salient feature of the app includes Gamification and Leaderboard. 

In this white paper, we explain the journey which was taken. How we decided on the resources and logic and to ensure costs are optimized on the cloud as they scale up number of users and consume more cloud resources. Attached PDF has more details

  • Scope and Objective
  • Requirement Gathering 
  • Approach
    1. POCs
    2. UI/UX
  • The Journey
  • Technology we used



  • Mobile app for Android and iOS platforms which can allow users to record Cricket bowling video to get bowling speed, line/length & trajectory
  • A monthly subscription service to get more detailed insights of the training sessions
  • Using ML libraries and using AWS Workloads to manage the intensive AL ML custom logic which was built

    We also used  AWS quicksight to report the Analysis and week on week user base statistics of how many active user exists, What device was being used etc. 

  • Custom logic build for AI & ML based video processing algorithm  to analyse the recorded videos and give consistent and accurate results
  • Serverless cloud platform which can scale to process hundreds of videos in parallel


  • Some issues we faced as below:
  • 1.  Accuracy of the result in some cases was not as good as we were relying on 1 camera, which in actual games are achieved using hawk-eye technique that involves 6.
  •     >> We added a feature on the app which informs the user if there has been some issue in calculating the accurate speed and gets highlighted with a red Mark, 
  • 2. Results were different as there were some variables example device and user ability to process videos,  different lighting conditions, different pitch.
  • >> We tried to train the model in different areas and time of the day to help ensure we can cover most scenarios and we added exception cases to show error messages to the user incase lighting was not good or if background had too much noise. 
  • 3. Provide a way to user to help them setting the phone correctly to minimize user error
  • >> An AR feature was implemented which guides user if the phone was not setup correctly. Training videos were released by the client on Youtube which was referenced on the app. 
Please fill in the details to download the case study.
[contact-form-7 id="3034" title="Case Study"]




Read More Case Studies