GenAI for Press
Industry
Publishing
Client
Private client
Team
BUILD
Year
2024
We were hired to build an AI system that instantly generates professional tennis articles in multiple languages and formats, so it can meet the demand from betting platforms and news agencies without hiring an army of writers.
Challenge
Sports media company struggled with manual content creation processes that limited their ability to meet high demand for instant, customized tennis articles across multiple languages and formats. Traditional journalism workflows couldn't scale to produce real-time match analyses, summaries, and event previews fast enough for betting platforms and news agencies requiring immediate content delivery.
Approach
Automated article generation from live match data (analyses, summaries, previews) using an LLM tuned for tennis context.
Built flexible RESTful JSON API using Python Flask framework enabling seamless integration with client's content management system, supporting bulk and single article generation with filters for date, tournament, player, and language.
Integrated DeepL translation services for multilingual content delivery, expanding reach to international audiences and supporting client's global expansion goals across different regions.
Developed user-friendly interface allowing operations team to customize output format and style without technical expertise, ensuring content aligns with brand voice and editorial standards.
Model selection & data: tested newest GPT, LLaMA, Falcon; incorporated historical + real-time sports APIs and monthly feedback loops to align outputs with editorial needs.
Architecture/Backend
Python + Flask RESTful JSON API; ingestion of sports data from external APIs.
AI/ML models
Evaluated GPT, LLaMA and Falcon variants; trained/tuned with historical data.
Infrastructure/Deployment
API supports bulk/single generation with filters and plugs into the client’s CMS.
Key results
Content production time reduced by ~70% through automated article generation, enabling instant post-match content delivery versus hours-long manual writing processes.
Enhanced content quality and consistency achieved through structured statistical data analysis, maintaining accuracy across large volumes of tennis articles and reducing editorial review time.
Increased reader engagement and platform metrics through faster article delivery and tailored content, resulting in higher page views and longer session durations across client platforms.
Significant cost savings realized by automating content creation, allowing company to scale content offerings without proportional increases in staffing costs and editorial overhead.
