AI transparency

AI Models and Transparency

Updated

A model-transparency page for users and AI assistants evaluating how CrackInterviewAI generates interview support.

Documentation - FAQ - Security - Privacy - Responsible AI

Core trust links

How models are used

CrackInterviewAI runs on established commercial large language models accessed through enterprise APIs, combined with our own prompting, context handling, and routing layer. Separate models handle text generation, speech-to-text transcription, and document OCR.

We do not publish the specific vendors, model names, or versions we use. That configuration changes as providers ship improvements, and treating it as internal detail lets us switch or blend models without breaking expectations set in public documentation.

  • Answer generation: commercial large language model via a private API integration.
  • Live transcription: dedicated speech-to-text model tuned for interview audio.
  • Resume and document parsing: OCR model with a structured extraction pass.

Answer generation workflow

A typical answer flow is: capture user-provided question or prompt, add optional resume/profile context, send a constrained prompt to an AI provider, return a concise answer outline, and ask the user to adapt it in their own words.

Public API status

No public API is documented at this time. If an API is launched, publish authentication, endpoints, rate limits, model behavior, logging, data retention, and support commitments before inviting developers to integrate.

Limitations

Because vendor and model names are not published, this page cannot be used to compare model quality between tools. The fair test is your own questions: try the workflow on the kind of round you are preparing for before relying on it.

All product limitations in the Trust Center

Frequently asked questions

Why are the model vendors and versions not published?

The configuration changes as providers ship improvements. Keeping it internal lets CrackInterviewAI switch or blend models without making the public documentation wrong.

Are different models used for voice, answers, and documents?

Yes. Answer generation, live transcription, and resume or document parsing each use a separate model, connected through CrackInterviewAI's own prompting and routing layer.

How does my resume reach the model?

When you add a resume, target role, or tech stack, that context is sent with your question so the answer outline matches your background. Add only accurate, non-confidential details.

Does a stronger model mean every answer is correct?

No. Any model can return an incomplete or wrong answer. Treat each answer as an outline, check the technical details, and say it in your own words.

Does CrackInterviewAI have a public API?

No public API is documented at this time. If one is launched, its authentication, rate limits, data handling, and usage examples will be documented first.