Privacy & Terms

Privacy

What we collect

Analytics. When you browse mllabs.com we run standard web analytics — Google Analytics, PostHog, Microsoft Clarity, and the LinkedIn Insight Tag. They record the pages you view, an approximate location (city or country derived from your IP address — never a precise position), your device and browser, and how you arrived (referrer and campaign links). This uses cookies and similar browser storage.

What you submit. Our forms and tools — the contact form, the ROI estimator's "save this estimate," the readiness scan's report unlock, partnership applications, and referral submissions — collect your name, work email, and whatever details you type into them. We use those only to reply to you, send the report or estimate you asked for, and follow up about the work.

What we do not collect

No accounts and no passwords — there is nothing to log into. We do not sell, rent, or trade your data, and we do not buy profiles about you. The readiness scan gives you a score without an email; you share one only if you want the full report.

Your choices

Analytics runs on an opt-out basis. Decline the cookie banner and we turn tracking off. Add ?notrack=1 to any URL to disable it permanently on your device, or block cookies in your browser. Want the details you submitted removed? Email [email protected] and we will delete them.

Terms

The tools are indicative

The readiness scan and the ROI estimator produce estimates from the inputs you give and the assumptions we state. They are a starting point for a conversation — not a quote, not a promise of results, and not financial, legal, or professional advice. A firm number comes from a scoping session and a written plan.

Ownership

The site, its copy, and the tools are the property of ML LABS LLC. The details you submit stay yours. Use the site for its intended purpose; do not scrape, resell, or misrepresent it.

Working with us

You own what we build

What we build for you is yours. On final delivery and payment, the intellectual property in the deliverables — code, models, and documentation — transfers to you. We keep only the right to reuse general know-how and non-client-specific tooling, never your data or your business logic.

How we handle your data

In an engagement we access only the systems and data the work requires, use them only to do that work, and return or delete engagement data on request when we're done. We do not train shared or third-party models on your data. Where you need AI that never leaves your walls, we build it on your own infrastructure with no outbound data flow.

Subprocessors

We build on standard cloud and AI infrastructure — major cloud providers and model APIs. The specific subprocessors touching your data are named in your written plan before work starts, so nothing is a surprise to your security team.

Procurement & insurance

We work through standard vendor onboarding. Business registration, a signed MSA/SOW, a mutual NDA, and a certificate of insurance are handled as part of procurement — tell us what your process needs and we'll provide it. Email [email protected] to start.

The guarantee, step by step

Fixed scope, written plan, measurable targets before work starts. Targets are written into the contract before work starts. Design and Build carry a full refund until acceptance — you sign when it's proven, or, once we've shown the targets met in writing, a 15-business-day evaluation window runs and closing it without naming a missed target is acceptance too. Either way you give up the system, not the money. Operate is different on purpose: no refunds, no lock-in — cancel any time with 30 days' notice.

The full terms

1. Written targets. Before work starts, the statement of work states the engagement's acceptance targets. Each target is objectively checkable: a named workflow, named data, and a pass condition a third party could verify. If a target cannot be written that way, it does not go in the SOW.
2. Acceptance. Acceptance happens one of two ways. Signed: you sign the acceptance certificate against the written SOW targets. Deemed: when ML LABS delivers written notice that the SOW targets are met, attaching the evidence the SOW names, a 15-business-day evaluation window opens; if the window closes without a written rejection naming a specific unmet SOW target, acceptance takes effect. Notice is delivered by email to the accepting officer named in the SOW and to the SOW's notices address; it is received on the earlier of written acknowledgment or the first business day after sending with no delivery failure, and the window runs from receipt, not from sending. A rejection that names no specific SOW target does not stop the window. No other act — requesting migration, using a staging deployment, going quiet before notice is delivered — constitutes acceptance.
3. Escrowed delivery (Build Tier I). ML LABS builds and operates the system in its own environment. You watch it run against the SOW targets, and you may migrate it to a staging or other non-production environment of yours to run your own security, latency, identity, and integration testing — doing so does not forfeit the refund right. Production data and production users come only after acceptance. Cancel any time before acceptance: all Build fees are refunded and the system stays with ML LABS — no license, code, or deliverables transfer.
4. Acceptance-anchored delivery (Build Tiers II–III). Where the system must be built in or integrated with your infrastructure, the refund right anchors to acceptance rather than location. All code and deliverables are licensed to you effective only on acceptance against the SOW targets. If you cancel before acceptance, all Build fees are refunded, the license never vests, and the system is withdrawn. For multi-workflow programs, acceptance and refund apply per workflow.
5. Regulated data. Where data may not leave your network, the build runs inside your VPC. Acceptance targets for regulated work are defined over de-identified or synthetic data — that is what the SOW targets name, so the guarantee is provable without production data. Validation against production data happens after acceptance, under the BAA or data agreement signed at SOW, inside the included Operate period; issues it surfaces are Operate work, not acceptance re-litigation.
6. Design. Design's acceptance target is fixed in the kickoff scope note, which names the workflow, the dataset, and the path the spike must demonstrate against that dataset. If the spike does not run as scoped, for reasons within ML LABS' control, the Design fee is refunded in full. The content of the recommendation — including a recommendation not to build — is not a target and not a refund trigger.
7. Scoping. The target is the deliverable: a written plan covering scope, delivery path, budget range, and a go / no-go. Not delivered, refunded.
8. Operate. Operate fees are not refundable. Either party may cancel with 30 days' written notice: the engagement ends 30 days after notice is received, the final period is prorated daily to that end date, and there is no cancellation fee and no minimum term beyond the notice period. The Operate period included in a Build — 30, 60, or 90 days by tier, as stated in the Build SOW — begins at acceptance; continuation past it happens only under terms agreed in the Build SOW, never by silent enrollment.
9. Adjudication. Whether a target was missed is determined against the written SOW targets, on the evidence the SOW names: logged runs, recorded outputs, and the pass conditions of clause 1. Before any refund claim, ML LABS has one cure period of 10 business days from written notice of the miss. If the parties still disagree after a joint review of that evidence, the dispute escalates: good-faith negotiation between principals, then mediation, then binding arbitration under AAA rules, venue Texas — a neutral arbiter neither party appoints alone, deciding on the clause-1 evidence. Misses caused by factors outside ML LABS' control — access not granted, data not provided, third-party outages, client-side changes — are not refund triggers; each SOW lists its client dependencies.
10. Credits. The $750 scoping fee is credited toward Design or any Build tier, and the $20,000 Design fee is credited toward any Build tier — both credits may be applied to the same Build. Credits expire 90 days after delivery of the engagement that generated them, are not transferable, and have no cash value. A refunded engagement generates no credit: any credit arising from fees later refunded is extinguished with the refund.
11. No reimplementation. While a system is in evaluation and for 12 months after any pre-acceptance cancellation, you may not use the delivered or demonstrated design, prompts, schemas, or evaluation approach to build or commission a substantially similar system. Accepting — and paying for — the work ends this restriction the moment the license vests.