Product descriptions:
SūmerSports sits at the intersection of elite sports decision-making and large-scale data science, serving professional sports teams and fans alike. With the rise of sensor-driven football data capturing player performance metrics, fragmented infrastructure and manual workflows slowed innovation and limited scale. Before Databricks, data engineering bridged disconnected systems with delayed insights and heavy overhead. With the Databricks Data Intelligence Platform, they built a unified data and analytics foundation that accelerated machine learning model development and deployment, enabled near real-time insight delivery to improve team performance and powered products like SūmerBrain to better serve their customers. What was once reactive is now fast, scalable and ready for the future of data-driven sports.
200 TB of player data inhibited scalable sports analytics
SūmerSports was created to help NFL and college football teams make better, higher-stakes roster decisions. As player contracts, draft picks, and trades increasingly carry multimillion-dollar consequences, the company’s mission is to bring clarity to those decisions by combining deep football expertise with advanced analytics. The emergence of large-scale player tracking data accelerated that opportunity. With sensors capturing player movement multiple times per second, teams suddenly had access to an entirely new class of data that had previously gone largely unmined. Founded in 2022, the startup was built to translate that data into something coaches, scouts, and executives can actually use.
To turn raw sports data into decision-ready intelligence, the company pulls from more than a dozen sources. This helps support player evaluation, trade analysis, draft strategy, and other critical football operations. “It really boils down to serving one of the highest quality datasets to our customers,” explained Matthew Coffey, Director of Data Engineering. “We currently have 14 different data sources that are getting aggregated into a premium data product.”
That foundation powers a suite of machine learning (ML) models that run continuously to deliver insights that shape decisions throughout the football calendar. “We have tons of jobs running nightly that are cleaning data, engineering features, retraining models, and serving inference on new data coming in,” said Adam Vonder Haar, VP of Data Science and AI.
SūmerSports has expanded those insights beyond professional teams to consumers through a fantasy football experience. The newly launched SūmerBrain is a football-specific chatbot that lets users interact with NFL data conversationally, syncing insights directly into their fantasy leagues. The same chatbot also exists in a team-focused version, allowing professional users to query data in a more intuitive, ad hoc way without being overwhelmed by complexity.
Behind these use cases, however, SūmerSports faced mounting challenges as the business scaled. The volume of data alone was significant. Each season, the startup processes hundreds of terabytes of structured and semi-structured data; sensor data, scraped web data, in-house scouting evaluations, client data, and financial information related to NFL contracts. “We’re probably processing up to 200 terabytes per season,” Matthew added.
As that data grew, the company’s early infrastructure struggled to keep up. Projects were fragmented and pipelines were inconsistent. There was no formalized way to manage lineage or orchestration across data ingestion, model training, and inference. “In a word, disorganized,” Matthew said of the pre-Databricks environment. “Every project had its own flavor. There was this constant worry of our data not being up-to-date or accurate.” Adam echoed that assessment, saying, “Visibility was limited. We only surfaced issues after they had already impacted downstream users.”
The underlying infrastructure compounded those problems. SūmerSports relied on a patchwork of Azure-native services, open-source tools, and custom orchestration that required constant attention from the data engineering team. Fine-grained control and governance were nearly impossible, forcing overly broad access patterns and heavy manual oversight. Without a centralized way to orchestrate workflows, manage dependencies, or trace how changes cascaded across models, data engineering became the bottleneck for the entire business. “Before Databricks, about 20% of my job was being the data librarian,” Matthew joked.
As SūmerSports expanded its customer base and added new products in its push toward real-time analytics, those challenges made it clear that the existing approach could not scale. The company needed a unified foundation capable of handling massive data volumes, coordinating complex ML workflows, and making trusted data accessible across the organization—without relying on constant manual intervention from engineering. That realization ultimately set the stage for their move to Databricks, which would become the backbone for how the company manages, analyzes, and delivers football intelligence at scale.
Democratizing analytics and ML development with Databricks
SūmerSports adopted Databricks as a single platform that now underpins nearly everything the company does with data. From ingestion and exploration to model development and customer-facing analytics, the Databricks Data Intelligence Platform provides a shared foundation that improves collaboration between data engineering and data science teams.
The Databricks Platform makes data broadly accessible across the organization. SūmerSports can now bring together sample data stored in AWS S3 to clean and transform it using standardized pipelines. Instead of siloed workflows and project-specific systems, teams now collaborate in a common environment where analytics, ML, and application development intersect. This provides greater transparency into how datasets are created, refined, and ultimately consumed. “Databricks has given us a central workspace for all the different functions to come together and have a shared truth,” said Adam.
A key enabler of that shared truth is Unity Catalog, which SūmerSports uses to govern access to their 14 primary data sources while ensuring business users work from the highest-quality datasets available. Rather than exposing raw source data, the team creates curated, tagged datasets that can be easily discovered and safely reused. According to Matthew, “Unity Catalog allows us to control access across different product teams while also providing lineage and search capabilities. We can securely expose the full picture of the data to analysts and data scientists so that they can do their jobs better.”
That governance layer reduces duplicated work and simplifies troubleshooting. Databricks Assistant, a context-aware AI assistant, is integrated with Unity Catalog. The Assistant removes the need for analysts to rely on data engineering to answer routine questions about tables and availability.
Databricks has also changed how teams explore and prototype new analytics products. SūmerSports makes extensive use of collaborative Notebooks and AI/BI Dashboards to research data, test assumptions, and iterate quickly on new reports and models. These visual dashboards make it possible to surface analytics to internal teams and customers before committing to fully productized applications. AI/BI Genie further accelerates this data exploration process, enabling business users to get instant answers in natural language rather than writing SQL.
For ML development, Databricks provides a standardized workflow that replaces what was once a scattered process. MLflow plays a central role by providing data scientists with the environment they need for experiment tracking and lifecycle management. “Being able to conduct exploratory data analysis and prototype different things in MLflow is a huge change,” added Adam. “Previously, this work lived on individual laptops, and knowledge was lost as projects evolved. Now it’s in one persistent location.” This standardization has reduced fragmentation and made it easier for new team members to understand how models were built and why certain decisions were made.
That consistency extends into production through the use of Asset Bundles, which SūmerSports uses as infrastructure-as-code. Instead of relying on a dedicated DevOps team to deploy and maintain infrastructure, data scientists and engineers can use shared templates to deploy their models to production themselves. “Databricks Asset Bundles allow us to deploy infrastructure-as-code for all of our model training, data cleaning, and inference jobs. This has changed how quickly ideas can move from development to production,” Matthew explained.
Databricks also supports SūmerSports’ generative AI initiatives. The startup has used Mosaic AI to prototype AI solutions and to create embeddings that power its chatbot interfaces, including SūmerBrain. While production models are currently served through AWS Bedrock, Mosaic AI provided a flexible environment for early experimentation and validation.
The Databricks Platform has simplified infrastructure management through built-in monitoring, reporting, and serverless compute. Alerts integrate directly with SūmerSports’ observability tooling, improving reliability without requiring custom monitoring systems. Perhaps most significantly, SūmerSports has shifted the majority of its workloads to serverless compute, eliminating the need to manage clusters for most jobs. According to Matthew, “About 95% of our workflow is on serverless. It’s one of the biggest boons in our day-to-day operations.”
Together, these capabilities have turned Databricks into a platform used daily across the organization. Approximately 50 technical users and 10 non-technical users regularly interact with the platform, querying data, exploring analytics, and validating insights themselves. “Databricks has become our best example of democratizing infrastructure at the company. Teams collaborate more effectively and focus on insight generation rather than tooling and maintenance,” added Matthew.
Delivering real-time intelligence at 3x the scale
With a unified foundation in place, SūmerSports has accelerated the pace at which new analytics and ML capabilities reach customers. One of the most visible improvements has been time-to-market for new models. Previously, moving a concept from exploration into a production-ready experience required months of coordination across engineering, data science, and infrastructure. Today, that cycle is dramatically shorter. Matthew described how a team-based Expected Points Added (EPA) model progressed from idea to real-time production. “We were able to serve customers that EPA model in just under four weeks,” he declared. “The same work would have previously taken three months.”
That speed has had a direct impact on SūmerSports’ ability to deliver insights when they matter most. One of the company’s core use cases is producing football AI outputs based on player tracking data shortly after games are completed. Early in the company’s journey, that process was slow. “In the first 12 months with Databricks, we cut that down from about four days to within the night. Sunday night games were available to customers by early Monday morning. In the last six months, we’ve cut it down to where it’s happening within the minute,” explained Matthew. SūmerSports delivers those insights in near real-time.
This improvement in speed to insight is foundational for both SūmerSports’ team-facing analytics and its consumer products, such as SūmerBrain. The ability to process, analyze, and surface fresh data quickly allows professional users to evaluate players, trades, and drafts with greater confidence. It also enables fans to interact with up-to-date NFL data conversationally. According to Adam, “This level of responsiveness has become a defining characteristic for us. What we hear most from customers is, ‘Wow, you guys move fast.’ We wouldn’t be able to enable this kind of speed—delivering player insights so quickly—without Databricks. It’s a core part of our competitive advantage, and Databricks has unlocked our ability to scale.”
The Databricks Platform has enabled SūmerSports to scale data processing alongside its growing customer base and product portfolio. As the volume and variety of football data increased, the company found it could process significantly more data without rearchitecting its workflows. When comparing a single week of NFL season processing to earlier systems, they are now processing nearly three times the amount of data in one week of the NFL season as they were two years ago. That increase supports richer analytics and more frequent model retraining. SūmerSports can expand into new use cases without compromising reliability.
Beyond measurable performance gains, Databricks has helped create a level of operational trust that allows teams to focus on innovation rather than oversight. In earlier environments, engineers and data scientists needed to constantly monitor pipelines to ensure they were running correctly. “We used to see every bit of data, and we had to see it because we couldn’t trust what was happening unless we had our eyes on it,” Adam recalled. With more reliable systems in place, that vigilance is no longer required, freeing teams to spend more time developing new capabilities.
Looking ahead, SūmerSports is using this foundation to push into more advanced and data-intensive frontiers. One priority is combining structured tracking data with unstructured sources such as written analysis and video. Adding video will paint a more complete picture of the game, enhancing SūmerBrain by enabling richer, more contextual information. The company is also focused on increasing analytical granularity. Historically, much of football analytics has operated at the play level. Adam sees the next breakthrough in understanding what happens inside a play. SūmerSports aims to uncover new insights into how and why plays succeed—insights that can inform both professional decision-making and fan engagement.
For a fast-growing startup, these advances are not isolated wins but part of a repeatable pattern. With standardized workflows and scalable infrastructure, SūmerSports is positioned to continue adding new data sources, launching new models, and expanding products like SūmerBrain without slowing down. As Adam summarized, the foundation now in place has “enabled the company to grow, move faster and explore opportunities that simply were not possible before.”
As the company moves forward, SūmerSports' vision is to redefine how sports organizations compete and win through the power of AI - elevating performance, sharpening execution, and delivering breakthrough results that revolutionize the game of football.