Talk
When your fintech doubles in size but your data team doesn't, you have two choices: burn out trying to keep up, or engineer your way out of the problem. We chose the latter. This talk walks through how a 4-person Data Platform team at Salmon Group, a BSP-licening digital bank in Manila. Built and operates a production-grade platform serving 20+ teams and 300+ data users across the business, without adding headcount. The Stack in Practice Our platform runs on Databricks on AWS, with a layered architecture from raw ingestion through bronze, silver, and gold marts via DBT. Data sources land through AWS DMS for change-data-capture ingestion, eliminating brittle custom connectors and enabling near-real-time availability of operational data. Jobs and workflows are managed as code using Databricks Asset Bundles (DAB), giving us reproducible, version-controlled deployments across dev, preprod, and prod, with PagerDuty-mapped alerting baked in from day one. DBT is the backbone of our transformation layer. Every model ships with unit tests, SQLFluff linting enforced at the MR level, and a GitLab CI pipeline that lets engineers validate changes in a preprod environment before anything touches production. Quality gates aren't an afterthought — they're the contract between the platform team and every squad that builds on top of us. The Cultural Lever The real scaling mechanism wasn't technical, it was self-service. We invested heavily in documentation, opinionated defaults, and tooling that makes the right thing the easy thing. Engineers across the business can run DBT models in preprod via a GitLab pipeline trigger, subscribe their jobs to PagerDuty alert tiers, and onboard new data sources without filing a ticket. This created a cultural shift: the platform team stopped being a bottleneck and became an enabler. Trust replaced gatekeeping What You'll Take Away Concrete patterns for building self-service data platforms at scale, CI/CD for data, infrastructure-as-code for orchestration, quality enforcement without bureaucracy, and how to cultivate a platform culture where four engineers can genuinely serve a company of hundreds.
About the Speaker
I'm a Head of Data Platform at Salmon Group, digital bank in Philippines, where I lead the engineering of a production-grade platform serving 300+ users across 20+ teams — running 500 daily jobs on Databricks and AWS with a team of four. I've spent over a decade building and scaling data infrastructure across fintech, ride-hailing, and telecommunications, at companies including Booking Holdings (Agoda), inDrive, and Megafon across Southeast Asia, Central Asia, and Russia. I speak regularly at data engineering conferences — Google DevFest Almaty, Kolesa Conf, Beetech, and Moscow Spark Meetups