Noor Alseginy
Back to the builds
Independent project · Deployed on AWS

Intake Review

Intake notes can contain missing or conflicting contact details. I built an AWS app that extracts the fields, shows their source, and requires a reviewer’s approval before a record reaches the CRM sandbox.

See how an intake is handled

These synthetic examples illustrate the workflow I tested in the deployed app.

Original intake note

Name: Avery Morgan
Email: avery.morgan@example.com
Phone: +1 202 555 0147
Request type: Consultation
Summary: Would like to schedule a consultation about a business contract.
After processing

Ready for review

The contact details are consistent and ready to check against the source.

Name
Avery Morgan
Email
avery.morgan@example.com
Phone
+1 202 555 0147
Request
Consultation
What happens next

A reviewer approves the fields, then chooses to export. Approval alone creates no CRM record.

What the live tests proved

I tested the signed-in browser workflow and inspected the saved DynamoDB records. The clean intake reached the CRM sandbox after approval and export. Conflicting phone numbers blocked approval. Uploading the same document again reopened its existing record.

How it runs on AWS
S3
Stores the uploaded document privately.
Lambda + Bedrock
Extracts fields and their supporting source passages.
Cognito + API Gateway
Requires a signed-in reviewer for upload, review and export requests.
DynamoDB
Stores review state and approved CRM records in separate tables.

The extraction worker and review API use separate IAM roles. The model prepares the fields. A reviewer approves them, and the API handles export.

A live browser test caught an API Gateway configuration issue. The browser’s preliminary CORS request was being asked for a token. I added an OPTIONS route for that check and verified that data requests still required authentication.