AI Dev Tools - Project Peer Review 2

We press on with the second of the three required peer reviews for the AI Dev Tools Zoomcamp certificate of completion.

Peer Review Number 2

This time, we review imend35's AI BAP - AI-Powered Business Analysis Platform . AI BAP connects "business requirements with reviewable User Stories, acceptance criteria, delivery tasks, and sprint planning."

I scored the project against the criteria found at AI Dev Tools Zoomcamp - Final Project .

The developer provided a pre-emptive FINAL_SUBMISSION_REPORT.md complete with scoring. Still, after personally reviewing the repo and live site, I asked my own agent to run its own evaluation, based on the code base and evidence provided. Unironically, my agent's score matched the developer's own final submission report, one point less than my hand-rendered evaluation.

Further Remarks for the Developer

2. Product Specification: 1 point. A specification exists (product-spec.md), but the 2-point level also requires a public task tracker and evidence that the agents used it. The pull-request numbering (#1 and #3 are PRs) suggests Issue #2 is the only issue. The developer’s own docs/github-issues-plan.md admits it was created after the work and cannot prove agents used a tracker.

6 Testing caveat: the “single command” is scripts/test-all.ps1. It is a PowerShell script that expects an isolated PostgreSQL test database already running on a specific port. That is stricter than “runs with a single command”, but the tests clearly exist and pass in CI, so 2 points is still fair.

![Expanded Frontend Automated Tests]https://cdn.bsky.app/img/feed_fullsize/plain/did:plc:gtlodux7dmvmo4xkiypm5b3q/bafkreihzp74urtvznx4gr33ajqz32o5mdhgxrp2tuj7xljtlj64d7asl3m()

Contradictory credential statement. The README says “No passwords or access tokens are published here” directly below a table that publishes two passwords. The demo accounts are low-privilege, but the author should fix one or the other.

Shared workspace. All reviewers write to the same data. The author warns about this, and it is a reasonable trade-off on a free tier.

Evidence style. Much of the README says what the “Product Owner reports” rather than showing it. The honesty is commendable, but screenshots or a short recording would make it easier to verify. Next step towards more points: the observability, alerting and on-call criteria together are worth 5 points. Instrumenting one user journey and adding one alert rule would be the most valuable addition.

On a personal note, the repo shows that the developer has a strong business analytics background, work experience, and / or education. It is commendable to take the step toward independent software developer, and I wish imend35 all the best.