Correct, on every product on this page. Everything I need is in the schema, the statistics views, the execution plans, your monitoring graphs, and for readiness work your dbt artifacts and pipeline run history. Your team can export all of it in a few minutes without granting anyone access to anything. You upload the files and answer a short set of questions, neither of which needs an account or a password. I deliver scripts and reports, and your engineers run them inside your own change process. If you later want me executing changes directly, that is a separate engagement with its own terms and its own price.

Every command is written out for you by engine, in copy-paste blocks. It is usually fifteen minutes for whoever already has psql or SSMS open. If pg_stat_statements or Query Store isn't switched on, I'll tell you how to enable it and what it costs you in overhead, which is close to nothing. If you truly can't produce them, we're not a fit yet, and I'll say so before taking your money.

Databases: PostgreSQL including RDS and Aurora, Microsoft SQL Server on-prem or Azure SQL or Managed Instance, MySQL and MariaDB. For readiness work, warehouses: Snowflake, Databricks, BigQuery, Redshift. I work on the data model, the queries, the indexes, the pipelines and the engine configuration. I don't take Oracle licensing work, Cassandra, or anything where I'd be learning on your budget. If you're on something not listed, ask. The answer is sometimes yes.

About fifteen minutes for a triage, an hour for a health audit, and two to three hours for a readiness assessment. That covers running the export commands, uploading the files, and answering the questions. Most of the readiness time is the nine questions about your AI initiatives rather than the commands. After that, nothing until delivery. No status calls, no weekly check-ins, no discovery workshop. If I need something I email you and you reply when you can.

TuBosun Abioye. Twenty years on databases, the last several as a database architect. I lead every engagement and write every finding myself. When the queue gets deep, overflow goes to senior engineers on my own team, not to offshore contractors and not to a junior running a script. Happy Database is operated by Candiid Labs Inc.

Different question, different exports, different buyer. The health audit asks whether your database is healthy. The readiness assessment asks whether your data can support the specific AI use cases you have already named. It reads dbt artifacts, warehouse metadata and ninety days of pipeline run history rather than execution plans, and it produces a score, a gap register and a ninety-day roadmap rather than a set of index scripts. If your AI work runs against a production database and you have no warehouse at all, there is a track for that too.

Against Gartner's five criteria for AI-ready data, so the framework is theirs rather than mine. Each of the five scores zero to four, weighted, giving a number out of a hundred and a band running from Not viable up to Production grade. Every level has an evidence test attached, which makes the score reproducible: someone else reading the same exports lands in the same place.

Four of the inputs are computed rather than judged. Your actual pipeline failure rate over ninety days, your documentation coverage at column level, your test coverage and how much of it actually blocks, and what share of your platform has no consumer at all. You also get a verdict per use case, because an aggregate score doesn't tell you which of your three initiatives is blocked and which is six weeks away.

Yes. There are three intake tracks. Track A is dbt plus a warehouse and gives the richest result fastest. Track B is a warehouse where transformations live in scheduled SQL or stored procedures. Track C is no warehouse at all, where the AI work runs directly against a production operational database, which is more common than people admit and usually the most urgent case. The score means the same thing on all three. Track A just takes me less time to reach it.

Most companies with an active AI initiative and no dedicated data platform team land between twenty-five and forty-five out of a hundred. That is the normal result, not a catastrophe, and I won't inflate a band to make the report pleasant to receive. The score is the reason you hired an outsider.

What makes it useful is the second half: which specific gap blocks which specific use case, in what order to clear them, what each one costs, and what score the roadmap should move you to. A low score with a costed ninety-day plan is a much better position than a good feeling and a pilot that has been six weeks from launch for two quarters.

No, and it is written into the privacy policy rather than just promised here. Your artifacts are never used to train any model, mine or anyone else's, and they are never submitted to consumer AI services. Where I use AI tooling in my own work, it runs under agreements that exclude inputs from training.

Worth adding that the question is narrower than it looks in this case. What you send is schemas, statistics and dependency graphs, not customer records. There is nothing in it a model could learn a person from.

Yes, but not on this page and not on a subscription. Remediation is a scoped project delivered by Candiid Labs, my engineering company, and the assessment roadmap is what it gets quoted from. The full $9,500 credits against it if you go ahead within sixty days. Plenty of clients take the report and hand it to their own team instead, which is a completely fine outcome, and the report is written to survive that.

You get a board. Add as many requests as you like. I work one at a time and deliver in about 48 hours, then pull the next. Bigger pieces like a partitioning strategy or a migration plan get broken down on my side and shipped in usable increments every day or two, rather than landing all at once at the end.

Nothing stays fixed, because you keep shipping. Every release brings new queries, new access patterns, and indexes nobody needed last month. That's what the monthly review covers. You export, I rank the top ten regressions against last month's baseline and fix them. If you hit a quiet quarter, pause the subscription. Unused days bank and wait for you.

Billing runs on 31-day cycles. Pause on day 21 and the remaining 10 days sit in your account until you come back. They don't expire. Cancel outright any time from the billing portal. No notice period, no exit conversation.

No. Every fixed-price product starts the moment you pay, upload your exports and answer the intake questions. Between them those do the job a discovery call usually does. If you want fifteen minutes first to check we're a fit, book it. I just don't make you sit through one to find out the price. For the readiness assessment most people do take the call, mainly to confirm which of the three intake tracks applies to their stack.

Usually not, and that is the entire point of the design. I hold no credentials, connect to nothing, and receive no production rows. Most security reviews exist to assess what happens if a vendor is breached, and there is nothing here to breach.

What normally satisfies a reviewer is one paragraph: the artifacts are structural and statistical, a schema dump describes tables rather than rows, an execution plan describes how a query ran rather than what it returned, and a dbt manifest is a dependency graph. Files go to a private upload link that shows the uploader nothing else, and they are deleted ninety days after delivery. Ask and I'll send that in writing the same day along with the privacy policy.

Yes. The card links are there because they're faster, not because they're the only option. If your finance team needs an invoice, a W-9, ACH details, or a PO number on the paperwork, email me and you'll have it the same day. Net 30 available on request for fixed-price work. Candiid Labs Inc. is a registered US federal contractor, so the vendor onboarding paperwork is already in order.

Yes, yours, at no charge, and it won't delay your start date. The terms already commit me to confidentiality, so an NDA is usually procedural, but plenty of companies need one on file. Send it with your intake and I'll return it signed.