Proof of Tech is a small engineering consultancy for teams shipping AI to production. For the systems that genuinely need it, we also build verifiable, on-chain inference. We're the team you bring in when the demo impressed everyone and now has to survive real users — and the team that gives you a clear go/no-go recommendation, in writing, before you commit to a build.
We are senior engineers who have designed, shipped, and scaled AI systems under production load. The team is deliberately small. We run one engagement at a time so principal engineering is never the bottleneck.
We publish under a single firm byline. Field Notes posts are written and reviewed by the engineers doing the work — we don't ship persona bios because the byline is editorial, not individual. Every post is sourced, qualified, and dated; the corpus is the credential.
We have spent years working across the surface area we now sell: production RAG, agent harnesses, voice systems on hard latency budgets, and verifiable inference where the math has to be checkable on-chain. The representative engagements on the homepage show the kind of problems we take on and the systems we build to solve them.
Every commit runs against a versioned eval suite. If the eval-pass rate drops below the budget set in week two, the deploy doesn't ship. The suite is yours at handoff — runner, dataset, scoring rubric, and the dashboard that watches it.
Two weeks in, you get a 12-page memo with a straight recommendation: build, don't build, or not yet. If the call is don't build, you keep the memo and the analysis behind it — a sound investment decision is the deliverable, every bit as much as a shipped system.
Code, evals, runbooks, dashboards — all of it. We retain no rights, no licenses, no required attribution. The only thing we keep is the right to reference the engagement publicly, with your written sign-off.
Most engagements won't need the on-chain piece. We build verifiable inference because some systems genuinely require it — gating governance, compliance proofs, federated eval. If yours doesn't, we won't pretend it does.
An engagement runs on a fixed cadence. Two weeks of diagnostic, six to eight weeks of build, and two weeks of handoff with a 30-day support window after. We embed inside your repo from day one, write the eval before the feature, and demo against the metrics — not slides — every week.
Pricing is a fixed fee for the diagnostic and a weekly retainer for the build. We bill outcomes, not hours. No success fees, no equity, no kickers. We do not subcontract — the engineers you meet in week one are the engineers in the repo in week six.
We run one engagement at a time. The calendar is the constraint, not the capacity. After handoff, the support window is the bridge; longer-running relationships are a separate retainer that we agree on in writing — see what handoff looks like for the details.
The engagements on the homepage are representative of the problems we take on and the systems we build for them. As work ships, we add named case studies — with the client's written sign-off, never before.
We follow CIS Benchmarks and operate inside your compliance perimeter when an engagement requires it. We don't currently hold SOC 2 attestation; where a regulation requires an attested vendor of record, we pair with one and say so in the engagement letter. Our full security posture is at /security.
References — current and recent — are available by name on a 30-minute call.
The first call is a free 30 minutes. You'll come away with a one-page memo on what we'd build and how we'd approach it.