SportsMax AI Widgets & Content Matching

The SportsMax widget is an embeddable, contextual AI product for sports publishers that reads and classifies each page's content, then dynamically generates the best experience, such as an AI-written poll or trivia question, using real-time sports data, partner insights, and imagery. Each experience is co-branded between a sports team and an advertiser, creating new ad inventory and extending teams' reach into the digital spaces where fans already are. As Lead UX Designer and Design Engineer, I own product strategy, UX/UI, and the technical build, including the article classification system, live-data integration, AI model testing and QA, and client relationships, taking the product from concept to near production-ready before handing off to engineering to scale.

Type

Type

Embeddable AI Widget

Embeddable AI Widget

Employer

Employer

Yieldmo

Yieldmo

Role

Role

Lead UX Designer &

Design Engineer

Timeline

Timeline

July 2025 - Present

July 2025 - Present

Problem: Publishers want to increase user engagement and time on page. Teams and advertisers want to extend their reach into new spaces and offer fans unique experiences. Delivering that at scale meant every experience had to be relevant to the page it appeared on, accurately and affordably.

Solution: I built a full pipeline that classifies each article (sport, topic, narrative), picks the right experience, like a fan poll or trivia, and generates contextual questions from live sports data. I validated classification with an LLM-as-judge process, tested models against cost, and created a token-based design system and publisher theming editor so widgets scale across sites.

Outcome: Model testing cut the full pipeline cost from $0.01 to $0.0008 per article (about 92%). At 20 large publishers, that's roughly $8 a day instead of $100, proving the product is viable at publisher scale. Working widgets are handed off to production, shaped by publisher feedback. 4 publishers are in the pipeline, and pilot launches will measure engagement and time on page.

Problem: Publishers want to increase user engagement and time on page. Teams and advertisers want to extend their reach into new spaces and offer fans unique experiences. Delivering that at scale meant every experience had to be relevant to the page it appeared on, accurately and affordably.

Solution: I built a full pipeline that classifies each article (sport, topic, narrative), picks the right experience, like a fan poll or trivia, and generates contextual questions from live sports data. I validated classification with an LLM-as-judge process, tested models against cost, and created a token-based design system and publisher theming editor so widgets scale across sites.

Outcome: Model testing cut the full pipeline cost from $0.01 to $0.0008 per article (about 92%). At 20 large publishers, that's roughly $8 a day instead of $100, proving the product is viable at publisher scale. Working widgets are handed off to production, shaped by publisher feedback. 4 publishers are in the pipeline, and pilot launches will measure engagement and time on page.

  • MITCH

  • KRUMM

  • MITCH

  • KRUMM