Data platform and lakehouse
Iceberg, dbt, Airflow, Spark. Stack assembled for the workload, not the trend cycle.
A data platform is six decisions: storage, format, compute, orchestration, semantic layer, BI. Get one wrong and you rebuild in a year.
I come in with hands-on stack assembly experience in a bank, fintech, and a product startup. I know the gap between Iceberg-in-docs and Iceberg-in-prod, between Airflow-with-2000-DAGs and Airflow-in-week-one, between “we'll take dbt Cloud” and “we run dbt Core plus our own scheduler”.
I design the stack so your team can carry it. Not the trendiest one — the one where the team goes home at night instead of into on-call. Every choice is captured in an ADR, so that a year or two later it's still clear why we picked it.
- For whom
- Companies with Postgres / OLTP in place who need a real data stack for analytics and ML.
- We solve
- BI runs directly against the prod DB and the product suffers. The data team uses notebooks as production. Every engineer picks their own tool. The cloud bill grows faster than the data.
- Deliverables
- An ADR per stack layer, a reference architecture with diagrams, a migration plan from the current setup, a 12-month operating cost estimate, and a PoC of a key pipeline.
- Format
- Remote.
- Duration
- 4–8 weeks.