Cutting DevOps overhead by 60% and transforming ML workflows

LSports is a global leader in real-time sports data, delivering unbeatable accuracy, reliability, and coverage to betting organizations worldwide. Backed by a powerhouse R&D team of 160+ developers, the company operates a data-intensive model, processing massive volumes of event-driven real-time data with ultra-low latency and high-performance scalability.

As LSports experienced rapid growth, so did the complexity of its data infrastructure, spanning PostgreSQL, MySQL, Redis, ClickHouse, Kafka, and more. In its pursuit of modernization, LSports embraced a data lakehouse architecture with Iceberg and Snowflake. But as data volumes increased and machine learning use cases expanded, the need for a powerful, cost-efficient, and easy-to-use query engine became urgent.

As LSports rapidly expanded and customer demands for real-time data products grew, the company looked to further evolve its data infrastructure. Key priorities included:

• Seamlessly accessing diverse data sources and Iceberg tables at scale.
• Minimizing operational overhead while ensuring consistent performance.
• Maintaining architectural flexibility to avoid vendor lock-in.
• Accelerating internal ML pipelines to support faster go-to-market for new features and models.

They needed a solution that could unify their architecture, reduce platform overhead, and scale with their data and AI ambitions.

LSports selected Starburst Galaxy after evaluating a wide range of query engines. Their team favored Starburst because:

• It offered Trino-as-a-Service with enterprise-grade SLA support and observability.
• It required zero manual setup to integrate with their Iceberg catalog.
• It eliminated the need to fine-tune infrastructure, saving platform and engineering time.
• It enabled internal ML workflows, such as data labeling and entity matching, with minimal latency and no vendor lock-in.

The first use case deployed was the LSports Simulator, which streams real-time events into its data lake and replays them for AI modeling. The team now uses Starburst as a backend engine for querying data used in ML pipelines, improving performance without relying on more costly or inflexible systems.

LSports was able to simplify data access across a fragmented architecture while accelerating critical machine learning and analytics workflows. The impact has been immediate, especially for the innovation and data science teams:

• Rapid deployment: Integrated with Iceberg and started querying data with minimal setup.
• Operational efficiency: Removed the need to manage Trino clusters and freed up engineering resources.
• ML acceleration: Enabled real-time query access for machine learning pipelines, boosting model accuracy and data access speed.
• Strategic agility: Supports multiple parallel engines like Databricks and BigQuery, ensuring flexibility in architecture choices.

These outcomes have reduced internal friction and paved the way for the broader adoption of Starburst across the organization. As LSports continues to scale its data strategy, Starburst is critical in building a high-performance, flexible, and future-ready data platform.