Challenge
Manual code reviews consume engineering time and can miss bugs, security risks, and logic issues across large pull requests.
t.
Opening workspace
loading portfolio modules...
All projects
Node 0107 Note 01 Note 02 Note 03
AI Code Review Platform
NitPick
An AI-powered developer platform that analyzes code and GitHub pull requests and turns them into structured code-review findings.
Context & approach
The problem, then the system.
Approach
Built an asynchronous review pipeline that routes code and pull-request diffs through FastAPI, Celery, Redis, and Gemini, validates structured findings, and stores review results for delivery to developers.
System map
How the pieces connect.
Explore the review pipeline. Select a service to inspect its responsibility, or replay a representative request path through the asynchronous AI workflow.
Request flow / illustrative
Move pointer to tilt · Select a node to inspectTap a node · Swipe to explore
Event source
GitHub
A pull request or review event starts the code-review workflow.
Engineering notes
Decisions that shape the experience.
Keep review work asynchronous
Celery and Redis separate code analysis from the incoming request path, with a Python worker handling queued review work.
Treat model output as an integration contract
Structured findings are validated with Pydantic before the review result is stored and returned to developers.
Persist the review result
PostgreSQL provides a durable place for review records after the analysis pipeline completes.
Built with
Technology stack
- Next.js
- Python
- FastAPI
- Celery
- Redis
- PostgreSQL
- Gemini
- Pydantic