Songtradr
Cloud-native music ingestion at a million releases a day
- Re-engineered release ingestion after our Kafka provider shut down and every alternative cost 2x or more. Moved it away from Kafka and lifted throughput from ~250k to 1M+ releases a day.
- Rescued a stalled search rebuild that had reached 30s+ per query, rewrote it on OpenSearch around the real requirements, and retired the legacy Solr service with sub-second, high-quality results.
- Built, migrated onto and ran a master metadata store, from tooling and feature work to ongoing maintenance. It took load off a legacy database costing ~$30k a month and rising, and now backs the platform's APIs.
- Ran 5–7 knowledge-share talks on adopting AI coding tools (risks, mitigations, tooling and good practice), starting with a group of ~10 peers and widening out across engineering.
- Contributed to security, platform design, support and data analysis across the wider engineering organisation.
- Worked XP-style with daily remote pairing.