Browse categories
Explore
Fiverr Pro
English
$
USD
I will audit your ai and llm app security for prompt injection and owasp risks
Bruno C
About this gig
AI features open new attack surfaces: prompt injection, data leaks through the model, and agents doing things they shouldn't. I'll audit your AI/LLM application and the web app and API around it, explain the risks in plain language and help you fix them.
I've spent 8 years building systems where security wasn't optional, including high-stakes platforms at XP, Banco Modal and LiveOn BaaS, and I build AI agents and RAG systems myself, so I know where they break.
What I check:
- OWASP Top 10 for LLM Applications: prompt injection (direct and indirect), sensitive data disclosure, insecure output handling, excessive agent permissions, system prompt leakage, unbounded consumption
- RAG and agent tool access: data isolation and least privilege
- OWASP Top 10 and API Security Top 10: auth, access control, injection, XSS, SSRF
- Secrets, dependencies and cloud/Docker misconfigurations
You get a prioritized report (critical to low) with evidence and fix guidance, plus the fixes implemented in Premium.
Strongest fit: Node.js/TypeScript, NestJS and React/Next.js. Ask about other stacks first.
This is a hands-on security review, not a compliance certification.
Get to know Bruno C
Bruno C
AI Engineer and Fractional CTO: Agents, RAG and LLMs in production
- FromBrazil
- Member sinceJun 2023
Languages
English, Portuguese
AI engineer and fractional CTO. I build production AI products: agents, RAG, LLM integrations and MCP servers, on the solid engineering that critical products demand. 8 years shipping software, including high-stakes platforms at XP, Banco Modal and LiveOn BaaS. I turn vague AI ideas into clear scope, honest estimates and on-time delivery, and I cover the whole stack around the model: Node.js/NestJS, PostgreSQL, React/Next.js, React Native, Electron, AWS/GCP with CI/CD and Docker, and OWASP security. I've led teams of up to 7 engineers as a technical PO.
Other AI Development Services I Offer
FAQ
What is prompt injection, and why should I care?
Text in user input or retrieved documents can hijack your model and make it leak data or misuse tools. It's the top risk for LLM apps.
Do you need source code access?
It's strongly recommended. A code + runtime review finds much more than black-box testing alone.
Will you test production?
Only with your written authorization. A staging environment is preferred.
Will you sign an NDA?
Yes, and findings stay confidential.
