Software development, cybersecurity, language models
We develop bespoke applications and platforms, assess the security of applications and infrastructure, and automate document processes. For each project, we define objectives, risks, responsibilities and acceptance criteria at the outset.
Working standard
In software projects, we document architecture decisions, access rights, behaviour under failure and operational responsibilities.
In systems based on language models, we define permitted sources, evaluation methods, limits of use and the process for validating results.
Our services
We select the architecture, security measures and language-model components according to the project's actual risks, data and requirements.
We assess applications, infrastructure and processes against the assets at stake, plausible threats and the real consequences of an incident. Each finding is tied to risk and to a verifiable corrective measure.
We develop applications and platforms around the organisation's actual way of working. Architecture, access rights, operations and maintenance are considered from the outset.
For systems based on language models, we define sources, access rights, evaluation and limits of use from the outset.
Cybersecurity
Before making recommendations, we identify the assets, attack surface, plausible scenarios and consequences of compromise. We then test the existing measures and prioritise remediation by risk, effort and dependencies.
The scope, assumptions and limits of the assessment are documented.
Access to the systems under assessment is limited to what is necessary and remains traceable.
Remediation is retested before a finding is closed.
Language models
Language models can produce fluent answers even when the available information is incomplete or wrong. We therefore define permitted sources, access rights, evaluation methods and how uncertain results are presented.
Sources can be identified and checked.
Access respects document and user permissions.
Evaluation begins with real tasks, not demonstrations.
How we work
We define scope, deliverables, responsibilities, acceptance criteria and risks before development begins. The project then advances through verifiable stages with documented decisions.
We establish the objectives, users, data, risks and acceptance criteria.
We define components, access rights, integrations and behaviour under failure.
We deliver incrementally, test the requirements and verify the relevant scenarios.
We hand over the code and documentation and establish monitoring, intervention and responsibilities.
Professional criteria
Scope, deliverables and important decisions are documented.
Complexity remains proportionate to risk, value and operational context.
Important changes, access and results can be traced.
Code, documentation and operations can be transferred to and sustained by the responsible team.
Direct contact
For an initial discussion, the context, systems involved, intended outcome and relevant deadline are enough.