Fixed-scope engagements

Most technical work is quoted after a discovery call, a scoping document and a proposal. That is three weeks before anyone writes a line of code, and it is how a small, well-defined problem ends up costing more to specify than to solve.

These three engagements are bounded on purpose. The scope is written down, the price is published, and the acceptance test is agreed before work starts — so you can decide in one reply instead of three meetings.

1. File conversion — US$850

You have an export from a system you are leaving, or a file in a shape nothing else will read. You need it in a schema something else accepts, and you need to trust it.

This one you can buy directly, without emailing first — the full scope and what is excluded are written out at /buy/file-conversion.

What you get:

• the converted file, in the target schema you specify
• complete source-row accounting: every input row either appears in the output or appears in a reject list with the reason it was rejected. The counts reconcile, or the run fails rather than quietly publishing
• a written policy for duplicates and for rows that are ambiguous
• one command that regenerates the whole output from the source
• tests you specify and can run yourself

Bounded at: one source format, one target format, up to 50,000 rows. Larger or multi-format work is quoted separately.

2. Integration triage — US$1,200, refunded if we cannot reproduce it

Something that used to work is failing intermittently, silently, or only in production. Before anyone should be paid to fix it, someone has to prove what is actually wrong.

What you get:

• a reproduction of the failure, or a written statement that it could not be reproduced and what that rules out
• root cause with the evidence that establishes it — logs, request bytes, timing, whatever actually demonstrates it
• a fix plan with acceptance criteria you can check
• a written risk note on what else the same cause is likely affecting

If we cannot reproduce the failure, this is refunded in full. We would rather return the fee than hand you a confident guess.

3. Reproducible test or evaluation harness — US$2,200

You have a system whose output quality you assert but cannot re-verify on demand — a model, a classifier, a matching or extraction step. You need a number you can defend, and regenerate next quarter without paying for it twice.

What you get:

• a scoring harness that runs against saved outputs, so a published result can be re-derived offline without re-running the system that produced it
• results that are byte-stable: identical inputs give identical output, every time
• explicit accounting for the awkward cases — abstentions, unparseable output, missing and duplicate records — reported separately rather than collapsed into one number
• a self-contained viewer, and documentation for your team to add cases without us

How payment works

We agree the scope and the acceptance test in writing first. You pay after the acceptance test passes, not before. For a first engagement, that means your risk is your own review time.

Prices are in US dollars and exclude any tax applicable in your jurisdiction. Larger work is quoted as milestones against the same principle: each one is defined, checkable and paid on acceptance.

What we actually work with

So you can rule us in or out without a conversation:

• Languages and runtimes: Python, JavaScript and TypeScript, Node, SQL, shell. Astro for static sites.
• Data work: CSV, TSV, fixed-width, JSON and XML in or out; schema mapping and conversion; reconciliation against a source of truth; deduplication with a stated policy; Excel output.
• Integration work: REST and webhook integrations, request signing, retry and idempotency behaviour, error handling and failure queues, API debugging against real traffic.
• Testing: automated test suites, reproducible evaluation harnesses, adversarial and failure-path testing.

Not in this list means we would be learning it on your budget, and we would tell you so rather than take the work.

What would make these a bad fit

Worth saying plainly, because a mismatch discovered later is expensive for both of us:

• If you need someone on calls — standups, live debugging sessions, stakeholder meetings — this is not that. The work is asynchronous and delivered in writing.
• If you need regulatory certification — accessibility conformance statements, security attestations, clinical or financial sign-off — we do not issue those.
• If the problem is not yet defined, a fixed price is the wrong instrument. Say so and we will scope it honestly rather than sell you a number.
• If you need a specific platform we have no track record in, we will tell you rather than learn on your budget.

Evidence, not adjectives

Rather than describe how carefully we work, the working samples are published with full source, their test records, and their stated limits. They are built on invented data and are demonstrations of method rather than client references — the pages say so themselves.

Toledo Technologies LLC is the contracting supplier. How we operate is set out on the about page.

Starting one

Email dev@toledotechnologies.com with the shape of the task. Useful things to include, if you have them: the source format, what the output needs to be, and how you would know it worked. Redacted samples are fine and often faster than a description.

You will get a written answer with a scope, a price and the acceptance test we would be held to. If it does not fit, the reply will say so and why — that is a normal outcome and it costs you nothing.