We build the systems
your business runs on.

Three shapes. Discovery is fixed price for two to three weeks. Well-specified work packages — an integration, a migration, a defined module — can be fixed price once discovery has produced the design. Ongoing product development runs as a dedicated team on time and materials with a contractual monthly cap, billed on named roles rather than a blended rate, so you can see what each person costs. Teams are typically five to eight people including a delivery lead and a QA engineer. Notice is 30 days for scale-down and 60 days for termination for convenience. There are no licence fees, per-seat charges or runtime royalties on anything we build for you.

You do, assigned on creation rather than on final payment, so a payment dispute never becomes an IP dispute. The contract schedules any pre-existing Atmora components used — mostly infrastructure modules and CI templates — and grants you a perpetual, irrevocable, royalty-free licence to them with no field-of-use restriction. Every release ships with an SPDX software bill of materials listing open-source dependencies and licences; we will not introduce copyleft-licensed code into your distributed products without written approval. We do not retain a copy of your source after transition, and we do not reuse client-specific code across accounts.

ISO 27001 and SOC 2 Type II, with reports available under NDA and a completed CAIQ on request. Engineers work on managed, encrypted, MDM-enrolled laptops; no client code sits on personal devices; access to your systems runs through your identity provider with hardware-backed MFA. Secrets live in HashiCorp Vault or your cloud's secret manager, never in repositories. CI runs SAST, dependency scanning and container scanning on every pull request, with builds failing on new criticals. Background and right-to-work checks are completed before onboarding. We notify you of a security incident within 24 hours of detection and provide a written root-cause analysis inside ten working days.

Systems deploy into your cloud accounts and your chosen regions — for UK and EU work that usually means London, Ireland or Frankfurt — and data does not leave them. Engineering staff in India reach non-production environments through a virtual desktop with clipboard and local storage disabled, so nothing is copied to a device outside the region. Production access is exception-based, time-boxed, approved by you and logged. Where regulation or internal policy requires it, we staff the engagement entirely from our London office at a higher day rate. Data processing agreements include EU standard contractual clauses and the UK international data transfer addendum, plus a sub-processor list with 30 days' notice of any change.

The statement of work names each engineer with role, seniority and allocation. We cannot rotate someone off without 30 days' written notice to you, a proposed replacement you can interview, and a two-week paid overlap that we do not bill twice. Key-person clauses cover the tech lead and the architect. Across the portfolio, 78% of engineers remain on an account for its full duration and average tenure at Atmora is 4.1 years. We staff from our own permanent employees; we do not backfill from contract markets when someone leaves, which is the usual cause of silent team churn.

Exit is designed in from the start: your repositories, your cloud accounts, your CI, no Atmora-hosted runtime and no proprietary framework, so there is no technical lock-in to unwind. Contractual transition assistance is 90 days at agreed rates and can be invoked for any reason, including a move to another supplier. Handover covers runbooks, architecture diagrams verified against the deployed system, recorded walkthroughs, a dependency and licence inventory, and paired on-call shifts with the incoming team. On request we run an exit rehearsal midway through the engagement: your team deploys and restores the system unaided while we watch. Anything it exposes is fixed as defect work at our cost.

Estimates come out of discovery with an explicit ±25% band and named assumptions; we do not quote precision we do not have. Anything that changes agreed scope goes through a written change request with its own estimate and an impact statement on date and cost, approved by your sponsor before work starts. Inside a fixed-price package we absorb our own estimation error — that is what the fixed price buys. Scope trades are encouraged: swapping work of comparable size within a sprint needs no paperwork. Every steering pack lists what moved out of scope that month and who decided it.

Four weeks of hypercare after cutover is included, with our engineers on your on-call rota alongside your team. Beyond that, managed support is a separate agreement with tiered response: P1 (service down or data at risk) 30-minute response and continuous work until service is restored, P2 four business hours, P3 next business day. Cover is business hours in your timezone by default, or 24x7 on a follow-the-sun rota across Bengaluru and London. Service credits apply against monthly fees for missed response targets. Monthly service reviews publish incident counts, mean time to restore, and the known-defect backlog with ages.

No. Everyone on your account is a permanent Atmora employee, which is what makes the continuity terms enforceable. We will not place contractors or partner-firm staff on an engagement without your prior written approval, and if you approve one they are named in the statement of work under the same background-check and device policies. Cloud providers and SaaS tools we use to deliver — source control, CI, observability — are disclosed as sub-processors in the data processing agreement, with 30 days' notice before any addition.

Yes, under enterprise agreements with zero data retention and no training on submitted content; we can name the vendors and share the terms during diligence. Generated code is reviewed by a human before merge and passes the same tests, SAST and licence scanning as anything hand-written — the author of the pull request owns it regardless of how it was produced. If your policy prohibits AI assistance, we disable it at organisation level for your repositories and can evidence that configuration. Where we build AI features for you, model choice, prompt versions, evaluation sets and refusal behaviour all live in your repository and are open to review.