Scoreline

Scoreline builds AI-powered fantasy sports products, live in the Premier League and the IPL. I joined to build its data platform from scratch and own everything from ingestion to the real-time pipeline that powers the app's live features.

Overview

Built Scoreline's data platform from the ground up, authoring over 95% of the platform's commits: eight ingestion sources, a five-layer BigQuery warehouse, and the real-time pipeline behind the app's live mode and live rankings. Alongside the platform work, I laid the groundwork for ML by researching and proposing the feature-store design and a leakage-safe gold feature layer.

Key Achievements

• Enabled the app's live mode and live rankings with a real-time event service offering replay-safe, idempotent delivery — zero downtime and zero data loss across every matchday.
• Laid the ML groundwork: researched and proposed the feature-store design, a leakage-safe gold feature layer, and a scoring engine that is 99.84% exact against official results.
• Hardened ingestion from vendor APIs, data scraping, and archives with idempotent loads, schema contracts, and cost controls, so vendor changes never break data or budgets.
• Established platform reliability: infrastructure and monitoring as code, 1,700+ tests, per-PR data environments, and measurement-driven alerting.

Technologies Used

Dagster, dlt, dbt, BigQuery, GCP, Pulumi, Datadog, and data scraping.