Built on a foundation of professional-grade hockey statistics and proprietary forecasting models. While they have a highly skilled internal engineering team, they recognized a massive opportunity to evolve: transforming their deep data into a conversational AI experience that allows fans and professionals to interact with advanced statistics across every major league (including the NHL).
To bring this vision to life, they needed a partner who could bridge the gap between their existing data and a high-stakes, user-facing AI, without disrupting their internal roadmap and missing a looming seasonal deadline.
Defined roadmap transforming sports data into products.
Built conversational AI delivering instant sports insights.
Developed scalable platform connecting data, AI, and users.
Enabled natural language queries for statistical analysis.
Most AI projects fail because the wrong problem was picked, not because the technology isn't available. We figure out which ones will actually pay off.
The primary requirement was non-negotiable: the system could not crash. Because this tool was designed for high-traffic environments, it had to be built to handle tens of thousands of simultaneous users. Stathletes needed a partner with a proven track record in building enterprise-grade, scalable solutions that remain stable under extreme pressure.
With the NHL hockey season fast approaching, Stathletes' existing engineers were fully committed to core operations. Redirecting them to build a complex AI product from scratch would have compromised their existing output. They needed an external partner that could hit the ground running without requiring internal management overhead.
While Stathletes are experts in hockey data, building a reliable SQL Agent that can accurately interpret natural language and query complex relational databases requires a specific set of AI disciplines. They determined that an external partner with deep LLM expertise was the most efficient way to fulfill this technical desire.
.webp)
We performed a deep-dive review of their existing data structures to build a solution that was both powerful and resilient.
We engineered a custom SQL Agent that allows the AI to "read" Stathletes' proprietary forecasting data. We mapped complex relational tables into a Data Dictionary, ensuring the AI pulls from the source of truth with 100% precision.
100%
Precise structured data.
We developed a classifier to distinguish between general hockey knowledge and proprietary data queries. This ensures the system remains fast and efficient, only querying the database when specific, deep-data insights are required.
To ensure the system remained professional and secure, we implemented multi-layered guardrails. These included boilerplate safety protocols (preventing off-topic subjects like politics) and specialized context-aware guardrails to keep the AI focused strictly on hockey analysis.
Using a scalable cloud infrastructure, we built the system to handle the Million User Test, ensuring performance remains lightning-fast even during the most intense periods of the NHL season.
~4M
Average audience reached during Stanley Cup broadcasts. 2025 season
By launching this AI analyst, Stathletes has opened a brand-new commercial channel. They now have a production-ready conversational product that can be licensed and customized to broadcasters, betting platforms, and professional teams.
We enabled Stathletes to develop a cutting-edge product at high speed without the risk and expense of hiring a new internal AI department. This allowed them to capture the market opportunity during the high season without increasing their long-term fixed costs.
Because NineTwoThree handled the heavy lifting of the AI development and scalability architecture, Stathletes' internal engineering team remained focused on their core business goals. The project moved from concept to launch without changing existing internal capacity.
The chat interface has created a continuous data pipeline of user engagement. By seeing exactly what fans and pros are asking, Stathletes has gained a new stream of business intelligence. This insight into market demand creates immediate opportunities for new data licensing and deeper audience engagement.
.svg)
NineTwoThree delivered a production-ready LLM chatbot on time, meeting the client's expectations. The team leveraged their experience to provide valuable inputs and stayed transparent throughout the process. Customers can expect a professional and trustworthy partner who follows through.
Build a new platform, or add AI capabilities to existing software, we deliver solutions that work in real operational environments.
Meet with founders Andrew Amann & Pavel Kirillov
Custom AI implementation roadmap
No sales pressure
