research_
active research projects:
- isolating AI models in self-improving sandboxes through red-teaming simulations, tackling various prompt injection attacks.
- replacing RLHF with intelligent middlewares for real-time filtering.