Data and Privacy
What Admit Architect stores, how it is encrypted, and our policy on model training.
Admit Architect stores the material you upload and generate for the purpose of running the agentic engine for you. We do not sell it, we do not use it to train any model, and we encrypt it at rest and in transit.
This page describes what we store, where we store it, and what you can do with it.
What we store
We store the data you give us and the data the engine derives from it.
Account. Your email, your authentication record, and your profile settings. Authentication is handled by AWS Cognito, so your password never touches our database.
Uploads. Any files you upload (resumes, transcripts, additional documents) are stored in a private S3 bucket. Extracted text from those files is stored in our database so the engine can reason over it.
Application state. Every application you create, every essay draft, every version history, every recommender invitation, every deadline, every connection, and every Mind session. This is the substrate the engine operates on.
Engine outputs. Dean Brief versions, narrative spine interviews, rubric evaluations, chances predictions, and every proposal Archer has ever made. All versioned, all auditable.
Where it is stored
Data is stored on AWS in the United States. The database runs on Amazon RDS with SSL enforced on every connection. File uploads live in a private Amazon S3 bucket with AES-256 encryption at rest. Model inference runs inside the same AWS account, so your prompts and responses never leave our environment.
Model training
We do not use your data to train any model, period. We also do not share your data with third parties for their model training. Every inference request runs in a mode that does not retain or reuse inputs for training.
Archer is an inference client, not a learning system. It gets sharper at helping you specifically because of what it can read from your Profile Memory, not because anyone is training a model on you.
What you can do
You can export the full state of your account (your applications, your essays, your spine, your dean brief versions) and you can delete your account and every associated record at any time from Account Settings.
If you delete your account, extracted text, interview answers, essay drafts, facts, evidence spans, and every session are removed. Raw file uploads in S3 are also removed. A short audit trail required for abuse prevention is retained and discarded after a standard retention window.
How it fits in the engine
Privacy is not a phase of the loop, it is a constraint on every phase. Ingest only writes to Profile Memory. Synthesize only reads from memory you own. Mind sessions are scoped to your account and are never visible to any other user. If a stage of the engine ever needed to share your data for it to work, we would tell you first and let you opt out.