AegisLabs · Personal project
Verifiable AI governance for NIST AI RMF, the EU AI Act, and OWASP LLM.
An enterprise-focused AI governance intelligence system. Engineers, compliance teams, product managers, and governance specialists get accurate, verifiable answers grounded in original framework text. Every claim ships with verbatim citations. The system abstains when evidence is insufficient.
Status
Prototype, deploying soon
- Domain
- Enterprise AI governance intelligence
- Stack
- RAG · Vector search · Citation validation · Risk classification · LLM eval · Python · Next.js
- Frameworks covered
- NIST AI Risk Management Framework · EU AI Act · OWASP LLM Top 10
Grounded in three governance frameworks
Core surfaces
Two purpose-built ways to use the system.
Surface · 01
Ask
Governance Q&A grounded in source text.
Plain-English questions return concise answers backed by verbatim quotes, article references, and direct source traceability. Honest abstention when evidence is insufficient.
- Verbatim source quotes
- Article and control references
- Confidence-aware responses
- Direct source traceability
Surface · 02
Assess
Structured AI risk and governance reports.
Describe a use case — model type, data sensitivity, user impact, autonomy, oversight, deployment context — and receive a full assessment.
- Risk classification (Low / Medium / High)
- Applicable governance controls
- Framework-specific obligations
- Recommended mitigation steps
- Evidence-backed reasoning with citations
Design principles
Three rules the system holds to.
01 · Principle
Verifiability over confidence
Every answer is grounded in actual framework text rather than unsupported model assumptions. Traceability and source verification beat polished AI summaries.
02 · Principle
Citation validation pipeline
Every quoted citation is matched against the original source document word-for-word before it reaches the user.
03 · Principle
Safe failure behavior
If retrieval evidence is insufficient, the platform abstains from answering instead of fabricating information.
Technical highlights
What's under the hood.
- Retrieval-Augmented Generation (RAG)
- Vector search and semantic retrieval
- Citation-grounded answer generation
- Verification and hallucination mitigation pipeline
- Risk classification workflows
- Governance mapping engine
- Evaluation and retrieval testing suite
- End-to-end AI governance assessment workflow
Skills demonstrated
What this project shows.
- AI Systems Architecture
- GenAI Engineering
- Retrieval-Augmented Generation (RAG)
- AI Governance & Responsible AI
- AI Risk Management
- AI Security Concepts
- Enterprise AI Solution Design
- LLM Evaluation & Verification
- Prompt Engineering
- Full-Stack AI Product Thinking
Project vision
Trustworthy, verifiable, and transparent AI for high-stakes work.
AegisLabs is a focused enterprise AI governance platform showing how AI systems can give trustworthy guidance for compliance and governance workflows. It reflects a broader interest in enterprise-ready AI that balances innovation with traceability and responsible practice.