AI Security
A collection of local educational labs designed to explore how modern AI applications can fail under adversarial input and unsafe trust assumptions.
- Python
- Ollama
- Streamlit
- FastAPI / Flask
- Docker
- ChromaDB
- LangChain
- SQLite
Work performed
- Prompt-injection scenarios
- Multi-level jailbreak challenges
- Poisoned knowledge-base experiments
- Model-cloning simulation
- Agent-compromise lab
- Malicious-plugin simulation
- AI attack-path analysis
- Full AI attack-chain simulation
Technical outputs
- Reproducible local labs
- Student exercises
- Setup and reset workflows
- Threat-model diagrams
- Assessment evidence
- Technical documentation
