AI-Powered Data Analytics For Sport Operations

The Client

Suppose you’re a sports enthusiast or, better still, a scout looking for a fast and interactive way to obtain reliable information about players, games, schedules, etc. You need insights into player performance, longitudinal trends, and statistics. This information will help you build rosters or choose which teams to support as a fan, so it must be accurate, factual, and up-to-date. Would you trust an LLM chatbot with this task?

Providing a dependable, domain-specific chatbot experience with inherently probabilistic LLMs under the hood is a challenge faced by one of our clients, a sports analytics company. We have successfully helped them bring their product from inception to production, and now it’s consistently serving over 50,000 prompts a month while processing more than 6 billion tokens.

To ensure output accuracy, we also developed an AI-driven engine that powers the chatbot’s analytical capabilities.

The Goal

Large language models excel at generating human-like text that looks plausible even when incorrect. An LLM-based service that provides data-driven analytics for high-stakes decision-makers like sports scouts must mitigate generative AI’s tendency to “hallucinate” with solutions that optimize for accuracy and reliability.

To that end, Janea Systems’ engineers set out to build a fast, secure, configurable, and trustworthy AI engine with deep analytical abilities to undergird our client’s chatbot.