01 How a project runs
Same destination. Different calendar.
Most of a long engagement is coordination: discovery workshops, handoffs, status decks. We removed the coordination and kept the engineering.
Why weeks
Four things make the difference, and none of them is working harder.
No discovery theatre
We start building in the first week. Understanding your data happens by looking at it, not by scheduling workshops about it.
No handoffs
One engineer and one AI. The person who scoped it is the person who builds it and the person you talk to.
A fixed scope, in writing
You know what is included before we start, and we tell you early if something changes it.
Visible every day
Work lands as pull requests in your repo. Progress is something you can read, not something you are told.
The phases
Illustrative shape of an engagement. Yours is scoped in writing before we start.
Week 0
Scope
A short call and a look at what you have today: sources, warehouse, reports, who uses what. You get a plain-English plan, a fixed scope and a date, in writing.
- plan
- fixed scope
- timeline
Week 1
Foundations
Repository, environments, access and automated checks come first, so everything after lands on rails. Infrastructure changes are planned and reviewed by a person before anything is applied.
- repo
- environments
- CI
- access model
Weeks 2 to N
Pipelines, models, reports
Sources flow, models are tested and documented, and the first reports show real numbers. You watch pull requests arrive in your repo as it happens; there is no status deck.
- ingestion
- tested models
- first reports
- run order
Hand-over
Yours
Documentation, decision records, runbooks and a recorded walkthrough. Your team runs it. We stay optional.
- docs
- runbooks
- walkthrough
- access cleanup
What we need from you
Access
Read access to the source systems and a place for the warehouse and infrastructure to live, in your accounts.
A decision-maker
One named person who can answer "which number is right" without a committee.
An hour a week
To review what has landed and tell us what is wrong. More if you want it.
What slows it down
Honest list: waiting for access to a source system, data that is far messier than described, and scope that grows mid-flight. We cannot remove those, but we surface them as soon as we see them, and we price scope changes instead of absorbing them silently.