I build the pipeline, ship the site,
and fix the data.

For an emergency — production down, a deadline today — usually within hours of your enquiry landing, because those are graded to the top of the inbox. For scoped work, a few days, depending on what is already in flight. You will be given a real date, not a "soon".

Because it decides the shape of the answer rather than the price of the hour. A small budget gets an honest "here is what fits inside it", which is a useful answer. A budget that cannot cover the work gets told so immediately, which is a better answer than a call that ends the same way an hour later. Nobody is asked for a figure they have not approved — it is a band, and "just exploring" is one of the options.

Yes, and it is a large share of the work. Half-finished repositories, a site that was built but never deployed, a pipeline whose author has left — all normal. The first thing you get is an honest read on what is salvageable and what is cheaper to rebuild.

Not to begin with. Most problems can be diagnosed from logs, an error message and a description of what changed. When access is genuinely needed, we agree the narrowest scope that works, through credentials you issue and can revoke the moment the work is done.

A working one, yes — extract, transform, load, and a dashboard at the end of it, running on your data rather than a sample. What a day does not buy is a hardened one: retries, backfills, idempotent reruns, schema-drift checks, alerting and cost controls come after, once you have seen it work and confirmed the shape is right. Building it in that order means you find out on day one whether the source data actually supports what you wanted — which is where most three-week discovery phases end up anyway.

That is the usual case, and it is the cheaper one. Whichever of Datadog, Grafana, Prometheus or the Elastic stack you already run, I work inside it rather than proposing a migration — a rip-and-replace mostly moves the problem to a new vendor and costs you your history. Two things get looked at first: which alerts fired in the last month without anyone acting on them, and which tags are multiplying your custom-metric count. Those two answers often cover the cost of the engagement on their own.

Either. Some clients want one specific thing fixed and then to be left alone; others keep a few days a month on retainer so the next thing has somewhere to go. You pick the shape in the first step of the form, and it changes what I quote.

Yes — most of it is. Everything is remote (email, shared repositories, screen share) and invoicing works in rupees or dollars. Time zones have not been a problem; overlapping hours are agreed before the work starts.

You are told that, and you pay nothing for the attempt. Where I can, I point you at who or what would help instead. Turning work down costs me one enquiry; taking work I cannot finish costs a reputation.