0 public items
External feedback starts at zero until linked GitHub issues prove otherwise.
8-minute public review
Help validate a data-quality LLM agent.
Try the public demo, inspect the support-ticket failure, and leave one GitHub issue with what was useful, confusing, or missing. Your review helps turn this project into public, resume-safe evidence.
0 confirmed
A confirmed user must state that they tried the demo or ran the repo.
26 issues ยท 1 fork
GitHub issue count includes public reviewer slots, feedback templates, and growth tasks.
252 tests
The repo verifies agent behavior, API routes, safety boundaries, and evidence artifacts.
Choose your review path
Pick the path closest to what you can honestly inspect. Each path opens a public GitHub issue template with permission and no-private-data checks so the evidence can be counted only when it is real, redacted, and non-owner.
Agent architecture
Inspect tool calling, planning trace, guardrails, structured output, and PostgreSQL boundaries.
Product feedback
Open the public demo and leave one useful, confusing, or missing workflow detail.
Run evidence
Run the local repo, container, CSV route, or PostgreSQL replay and submit observed output.
Real workflow fit
Share an anonymized data-quality scenario, expected impact, and whether the agent maps to it.
What to check
The demo analyzes an anonymized support-ticket export. The expected result is a failing quality report with duplicate IDs, missing routing fields, negative amounts, and outlier evidence.
1
Open the public demo
Skim the report snapshot and confirm whether the data-quality failure is understandable.
Open Demo2
Check the AI-agent signal
Look for tool-calling, read-only data boundaries, structured evidence, verifier guardrails, and fallback behavior.
Agent Capability Matrix3
Leave one public note
Write what worked, what confused you, and whether the project feels credible for an AI Engineer internship review.
Submit GitHub FeedbackCounts toward
external_feedback_items.4
Optional: confirm a run
If you ran the local repo or replayed the CSV/PostgreSQL path, submit a redacted replay note.
Submit Replay EvidenceCounts toward
reproducible_feedback_items.Safety boundary
Please do not post raw customer data, secrets, private emails, addresses, tokens, production rows, or private company names. Redacted summaries are enough.