— 01
Grounded answers.
IRIS answers only from your institutional documents, never from the open internet, never from model memory. Citations on every reply, linked to the source page.
Memora Product 01 — IRIS
IRIS.
A secure AI assistant for institutional research and knowledge management. Live at NJIT today.
Actual response from IRIS at NJIT, grounded in the live Common Data Set knowledge base.
Section 01 — What IRIS does
Four things,
done quietly.
— 02
Conversational.
Plain-language questions. No SQL, no dashboards, no training required. The first question is always the easiest one to ask.
— 03
Secure by design.
Single-tenant deployment in AWS, operated by Memora. Your documents stay inside the AWS network and are contractually never used to train any model.
— 04
Fast to deploy.
Two to three weeks from document handoff to production. Most of that time is yours, not ours. Your team approves the answers IRIS is allowed to give.
Section 02 — Try it
Ask a real question.
Below is IRIS connected to NJIT's institutional research corpus. Click any sample question; the answer streams in, grounded in the documents it came from.
Source documents include the Common Data Set, Factbook, Cookbook, methodology notes, and the Faculty Handbook.
Section 03 — Architecture
Under the hood,
five steps.
A retrieval-augmented architecture, built on AWS-native components. Technical enough for a CIO; legible enough for an IR director.
Section 04 — Security & data
Security is not a feature. It's a foundation.
IRIS is built for the institutions that have to be right about who can see what. Single-tenant deployment inside AWS, operated by Memora, with contractual guarantees that no client data is ever used to train any model.
— Infrastructure
AWS Bedrock + OpenSearch Serverless + Lambda + Cognito.
Single-tenant resources per client, hosted in AWS us-east-1. We handle all provisioning and operations inside our AWS environment. Your team connects via SSO; your documents stay in AWS and never leave the network.
— Compliance
SOC 2, ISO 27001, FedRAMP High: at the platform level.
Memora inherits SOC 2, ISO 27001, and FedRAMP High compliance posture through AWS Bedrock. Memora's own certification roadmap will be pursued as client requirements demand.
— Training
Your data is never used to train any model.
AWS Bedrock enterprise terms contractually prohibit using customer data for training. No data leaves the AWS network.
— Scope today, scope ahead
Public & internal documents now. Regulated data on the roadmap.
FERPA, HIPAA, and PII-bearing corpora are on the roadmap when client requirements and our compliance posture support it.
Section 05 — Deployment
Three weeks from handoff to launch.
— Week 01
Ingestion & setup.
We work with your IR team to assemble the document corpus. We provision a single-tenant AWS stack, with SSO connected to your identity provider. The knowledge base is loaded by end of week.
— Week 02
Tuning & evaluation.
We assemble a private test set with your domain experts. IRIS is tuned against the questions your team actually asks. We measure faithfulness, citation accuracy, refusal behavior.
— Week 03
Pilot & launch.
A small pilot group inside your IR office. Daily check-ins with your team. By the end of week three, IRIS is live for the broader staff.
Next step
See IRIS in your
own environment.
A thirty-minute conversation with our founder. We'll talk about your institution, your documents, and what a deployment would look like, before anyone draws a contract.