Infrastructure that keeps
your data moving.
We design and operate data pipelines, cloud platforms and observability for teams of five to fifty. Small studio, senior engineers, no account managers.
Three things we are good at
Most engagements start with a two-week audit and a written plan. If the plan is good, we usually build it too.
Pipeline engineering
Batch and streaming pipelines on Kafka, Airflow and dbt.
- Backfills without maintenance windows
- Schema evolution and replay
- Cost-aware warehouse modelling
Cloud platforms
AWS and GCP landing zones that survive an audit.
- Terraform modules and state layout
- IAM and network segmentation
- Cost and quota guardrails
Reliability
SLOs, alerting and runbooks for the systems we build.
- Error budgets that mean something
- On-call handover and drills
- Postmortems without blame
Selected work
A logistics platform ingesting 400M events a day; a fintech ledger moving to a streaming model; a research group that needed reproducible pipelines on a small budget. Case notes available on request.
Latest writing
-
Schema evolution without the fear
How we move a warehouse table from one type to another while six teams keep reading it.
-
Running Airflow on two vCPUs
What actually breaks when you do not have a scheduler farm, and the knobs that matter.
-
Backfilling a billion rows without a maintenance window
Chunked replays, idempotent writes, and the one metric we watch while it runs.