AWS for M&E Blog

How Infosys reimagines the game of tennis using AWS

This post has been co-authored by Rohit Agnihotri, Chief Architect of Infosys Tennis Platform, and Satheesh Kumar, Principal Solutions Architect at AWS

Infosys is a global leader in next-generation digital services and has developed innovative solutions to engage stakeholders in the sports ecosystem. Infosys is the official digital innovation partner for major tennis events like the Australian Open, Roland-Garros, ATP, and the International Tennis Hall of Fame. Infosys has transformed the way tennis games are viewed and analyzed, enabling fans, players and coaches, broadcasters, journalists, and organizers of tennis tournaments to experience the game with a new set of immersive and intelligent tools that aid in diving deep into the sport.

This blog post describes game-changing innovations powered by Infosys and how the Infosys Tennis Platform uses Amazon Web Services (AWS) to enable digital transformation.

Solution overview

The Infosys Tennis Platform powered by Infosys Cobalt is available in two variants 1) Serverless and 2) Cloud Neutral. This blog post focuses on the AWS-based serverless stack where the Infosys Tennis Platform relies on managed services from AWS to achieve operational excellence, security, reliability, performance, and cost efficiency.

Media pipeline

While multiple feeds are ingested into the Infosys Tennis Platform solution, at its core is a media pipeline that processes live broadcast feeds using AWS Media Services and custom machine learning (ML) models to bring out interesting insights from tennis matches.

A live broadcast feed from the tennis court is shared as RTMP input to AWS Elemental MediaLive. The MediaLive channel processes the live feed and breaks it into smaller chunks of compressed video files and stores them in an Amazon Simple Storage Services (Amazon S3) bucket. These smaller chunks of video files, along with statistical data, are used to generate video clips using AWS Elemental MediaConvert.

These video clips then pass through a series of Infosys proprietary artificial intelligence (AI) and ML models to pick up insights such as crowd excitement detection and Hawk-Eye reviews. This helps in creating AI Highlights & AI Videos. Custom ML models run on AWS Lambda using container images stored in Amazon Elastic Container Registry (Amazon ECR) and generate the relevant metadata that gets stored in an Amazon Aurora Serverless Postgres database. The processed data is exposed via Lambda APIs that, for example, allow coaches to slice and dice videos and generate custom video clips to share with players. The use of AWS managed services allows for processing of live broadcast feeds and makes the game analysis video available in a near real-time fashion.

In this diagram, two types of feeds are processed. First, the live RTMP feed from the game court is processed by the media pipeline using AWS Elemental MediaLive and MediaConvert services to create video clips. Second, Hawk-Eye and statistical feeds are pushed into Amazon Kinesis streams. Lambda processes these events and stores them in Aurora Serverless Postgres database.

In addition to live broadcast feeds from matches, scoring, statistics, player position, and ball tracking data are ingested into Amazon Kinesis data streams in near real-time using a TCP server running on AWS Fargate. AWS Lambda is used to process these incoming data feeds before storing them in an Amazon Aurora Serverless Postgres database. The pipeline used to ingest feeds can scale easily to accommodate more data feeds from multiple tennis courts.

With Amazon Kinesis data streams, one can take advantage of enhanced fan-out so that consumers can process with increased throughput of up to 2 MB of data per second per shard. This comes in handy to ingest multiple game feeds at a faster pace. Currently Infosys Tennis Platform scales to handle close to one million events during a major tournament.

With Lambda consuming events from Kinesis streams, the parallelization factor can be set to control the number of concurrent Lambda invocations for each shard in Kinesis data streams. We use a concurrency factor in the range of 40-50 on multiple Lambda functions to process incoming events.

Aurora Serverless Postgres DB automatically scales up and down depending on the number of required read and write operations. With the entire solution stack on AWS being serverless, the operational overhead of scaling and managing workload is reduced significantly.

The user experience

To provide interesting insights about a tennis match, the Amazon S3-hosted web application pulls the required data using REST APIs written in AWS Lambda and exposed via Amazon API Gateway. As these are internet facing services, Amazon CloudFront is used as the entry point with AWS WAF, providing the layer-7 protection.

This solution also ingests Twitter feeds to capture game-related tweets and uses a bespoke Spark application on Amazon EMR to process tweets and perform sentiment analysis using Amazon Comprehend. The results are indexed using Amazon OpenSearch. Insights can be showcased on a variety of dedicated user portals to better engage fans.

This diagram depicts how REST APIs deployed as Lambda functions are exposed via Amazon API Gateway. A frontend static page in Amazon S3 is rendered via Amazon CloudFront. The diagram depicts how game-related Twitter feeds are ingested into Amazon Kinesis streams and processed using Amazon EMR and Amazon Comprehend services for sentiment analysis of the tweets.

The stakeholder experience

The Infosys Tennis Platform can change the way stakeholders experience tennis in interesting ways.

Players and coaches

The output of the media pipeline previously described are the videos generated via the Infosys AI videos solution, which offers AI-powered match analysis with cognitive capabilities such as natural language search. The solution can instantly generate insights about strengths and areas of improvement for players in a match. Insights help players and coaches make informed decisions about training and match strategy.


Infosys Match Centre provides interactive data visualizations of game stats and 3D Court Vision providing a closer view of key events. Fans can enjoy live commentary for each match, describing the point and the score, as well as key metadata such as speed of serve and length of rally. Game insights extracted from video clips using custom ML models are available to fans, allowing them to immerse themselves in the game by following the ball trajectory for each point as it’s played.

Infosys Tennis Platform provides an immersive experience through 3D Art Museum for digital exhibition of artefacts associated with the sport of tennis. This component is developed using Unity 3D, Maya and Babylon.js, hosted as a static website using Amazon S3.


Infosys AI highlight videos can be useful to journalists covering the game of tennis. Video provides instant access to key moments in a match including crowd sentiment, statistical data, and other ML-powered insights. These highlights can be sized for online posting to social media platforms without editing.


The Infosys Tennis Platform demonstrates how a scalable digital solution on AWS can enrich the game experience and add value to the sports ecosystem. For more information about the Infosys Tennis Platform, please click here. And to get more information on Infosys Cobalt, please click here

Rohit Agnihotri is the Chief Architect
of Infosys Tennis Platform.

Satheesh Kumar

Satheesh Kumar

Satheesh is a Principal Solutions Architect at AWS. He works with enterprise customers to help them build solutions using AWS services.