The practical details, explained plainly.
Understand fit, data handling, evaluation limits and project arrangements before you share access or commit to a scope.
Where to begin.
Where do we start?
With one workflow and a decision you need to make. Describe what the system should do, what goes wrong and what you already have. We’ll discuss a suitable scope before requesting access.
Can you build the AI application as well?
Acadify AI focuses on evaluation, datasets and training. For application development and integrations, Acadify Solution’s engineering practice is the connected route.
Do we need a complete test dataset?
No. Describe the assets you have. Preparing references and defining task coverage can be part of a separately agreed scope. Without reliable references, some conclusions may remain uncertain.
What does an evaluation cost and how long does it take?
Cost and timing depend on task coverage, data readiness, execution access and review requirements. They are confirmed in a scoped proposal; there is no universal price or delivery promise.
Can you compare commercial and open models?
We can discuss candidates that your access, licensing and provider restrictions permit. A useful comparison uses the same tasks and recorded conditions, including cost assumptions and response-time measurements.
Agree the boundaries first.
How do you handle private data?
Describe the problem without sending private records. During scoping, agree data access, permitted providers, retention and deletion. Ask about an NDA before sharing proprietary material. Any client-hosted execution needs a technical review.
Can we require our own environment?
Raise that requirement during scoping. Client-hosted execution depends on the technical setup, available access and agreed controls. It is not automatically included in every engagement.
Who owns the report and test assets?
Ownership, permitted reuse and any third-party restrictions are specified in the engagement terms. Confirm them before starting; this page does not replace a contract.
Should we send customer records through the contact form?
No. Start with a description or a redacted scenario. Arrange an approved transfer method after access, confidentiality and retention requirements are agreed.
What a result can tell you.
Will you certify our AI as safe or compliant?
No. Evaluation provides evidence about the tested tasks and conditions. It does not certify legal compliance or guarantee error-free behaviour. Your accountable owners decide deployment and review requirements.
Are the public samples real client case studies?
No. The policy sample uses fictional data and hand-written candidate records. It demonstrates exact checks and reporting, not real model accuracy or delivered client results.
Will a passing result guarantee future behaviour?
No. A result describes the tested versions, tasks and conditions. New documents, prompts, permissions or model versions can change behaviour. Keep a regression baseline and review coverage gaps.
Can we work with you from outside India?
Enquiries from international teams are welcome. We agree working-hour overlap, communication, access constraints and commercial terms before committing to an engagement.
Have a question about your own workflow?
Share the workflow and the problem. We’ll confirm fit, then discuss scope, access, deliverables and cost before work begins.
Discuss your project