How we work
Start with the question you need answered or the software you need built. We agree on the scope, success criteria, and responsibilities before work begins.
A written review of an AI idea, an existing product, agent readiness, or a data problem, with a clear recommendation.
For decisions that need evidence, a technical review, and a clear recommendation: build, simplify, defer, or stop.
Usually 1 to 3 weeks for a focused scope
Compare assessmentsWe turn a defined idea into working software your team can try.
For a clear user need that a small piece of working software can test.
A prototype may be ready in days, depending on the task and access
See prototype developmentWe design, test, and deploy a product, internal tool, automation, integration, or data system.
For a defined first release with a clear owner and a known place to run it.
A focused release can ship in weeks when access and decisions are ready
See software development and integrationWe coordinate several related projects through a shared architecture, agreed priorities, and staged releases.
For products, processes, or data systems that need to change together.
A staged program with early releases
See transformation programsKeep fixing what slows users down. We turn feedback and recurring problems into a prioritized development plan and tested releases.
For software that needs continued attention rather than a final handoff.
A regular review and release schedule
See ongoing improvementWe take on agreed responsibility for monitoring, maintenance, review, and reporting.
For teams that need clear coverage and escalation after release.
Ongoing, with coverage set out in the agreement
See managed operationsThe Permadyn delivery standard
The release includes the agreed code and configuration, test results, known issues, and instructions for the team running it. We define the required checks and who accepts the result at the start.
See the release review and download the checklist
We write down the workflow, deliverables, current baseline, exclusions, and the person who accepts the result.
Before building, we confirm which data sources are permitted, where data can move, what access we need, and which model or infrastructure providers are involved.
You review representative data or a working part of the application. We settle quality and integration questions before the scope grows.
We run the agreed checks for correctness, exceptions, permissions, and reproducibility. Model behavior is tested separately from ordinary software correctness.
We deliver the agreed source code, infrastructure configuration, evaluation results, deployment instructions, and recovery procedures. You confirm the release, and we document known limitations.
We agree how corrections and changes are handled, who maintains the system, and any support coverage you contract with us.
Begin with a readiness review, deploy one focused application, or build a shared internal AI platform in stages.
We plan the hardware, model choice, integration, rollout, maintenance, and the tests that confirm data stays inside your boundary. The first conversation needs only a non-sensitive description; access to protected material is agreed separately.
See on-premises AIA focused data engagement
Get realistic, fictional records for one development, QA, or staging workflow. We start from your schema, the relationships that must hold, and the exceptions your application has to handle.
We confirm source requirements, integration effort, record volume, and acceptance checks before pricing the work. The pilot does not need production customer records if the schema and business rules are enough.
Scope a data pilotInspect the engineering referenceWe define the scope, deliverables, acceptance criteria, and price with you. The agreement also covers source-code rights, account ownership, third-party costs, and any support. Timing depends on access, integrations, security requirements, and how quickly decisions are made on your side. The first review may recommend a smaller project or conventional software.