Previous Case Study • Fintech / Developer tools
FakePE
API sandbox enabling developers to test edge-case payment flows.
Case FPRead full study →
Back to projects
AI / Developer tools
CodeDog
AI-assisted security reviews directly in developer pull requests.
< 12s
Scan Time
-92%
False Positives
84%
Remediation Rate
codedog.velnix.studio
pull_request_review_7.ymlSecurity Pass
@@ -42,8 +42,8 @@ function handleInput(req)
- const query = `SELECT * FROM users WHERE id = ${req.body.id}`;
+ const query = `SELECT * FROM users WHERE id = $1`;
CodeDog Security Agent Analysis:
"Identified SQL Injection vulnerability. Direct string interpolation into DB query exposes the system to query spoofing. Refactored query to use parameterized query parameters instead."
Overview
CodeDog bridges the gap between complex static analysis and active developer workflows. By running automated scans on incoming Git diffs and explaining vulnerabilities in plain language, it turns security checks from an engineering bottleneck into a collaborative step.
The Stack
ReactTypeScriptPythonFastAPIOpenAI APIDocker
What We Shipped
Real-time secure LLM parsing engine
GitHub Actions & GitLab webhook receivers
PR comment synchronization pipeline
Interactive vulnerabilities status dashboard
The Challenge
Traditional security scanners output hundreds of false positives, which developers routinely ignore. Generating static reports outside of GitHub or GitLab workflows means vulnerability remediation is delayed until just before shipping, causing major design delays.
Our Solution
We designed and built a secure, real-time LLM-powered review pipeline. Instead of bulk reports, CodeDog acts as a junior reviewer on PRs—analyzing the delta of the code, identifying actual vulnerability paths, and providing direct, copy-pasteable diff corrections in comments.
Keep Exploring
Other Case Studies
Next Case Study • Infrastructure
VedDB
High-performance key-value database designed for operational clarity.
Case VDRead full study →