AI security engineering
Mallace: Security fixes backed by proof
AI-assisted vulnerability discovery, trusted regression tests, isolated repairs, and independently reviewed pull requests.

The product
Mallace connects security discovery to a reviewable code change. It scans a repository, asks a model to produce a regression test for a suspected vulnerability, and requires that test to fail on the current code before attempting a fix.
Prove before repairing
A proposed repair must make the same test pass. Docker isolation keeps test execution separate from the host, and an independent model reviews the change before delivery. One pull request per verified fix keeps the evidence and implementation together for human review.
A complete product
The Python tooling and GitHub Action connect to a Next.js product interface and a FastAPI/PostgreSQL license service. Billing, usage budgets, and execution boundaries belong to the workflow alongside the model integration.
Engineering focus
The important boundary is between a plausible model response and an observed result. Tests, sandboxing, independent review, and a human merge decision make that boundary visible. Mallace is an alpha product with explicit supported checks and repair paths.
Stack
- Python
- FastAPI
- Next.js
- PostgreSQL
- Docker
- GitHub Actions