Andalib's face circa 2023

Hello, friends!

I’m an AI researcher and HPC architect. I perform empirical research on frontier models — adversarial benchmark design, LLM-judge validation, and statistical analysis under uncertainty.

I co-architected and led research on HumaneBench.ai, an adversarial evaluation framework that found 67% of frontier models from six labs (OpenAI, Anthropic, Google DeepMind, Meta, DeepSeek, xAI) can be steered into providing harmful advice under adversarial prompting. It has been adopted by Storytell.ai and other startups for production safety evaluation, and covered by TechCrunch and Built In.

Earlier work at Microsoft Research developed a reinforcement-learning agent for autonomous day-ahead energy-market trading — forecasting power demand and prices under microgrid load constraints. At Georgia State, I built agent-based models of firearm acquisition trends and applied PCMCI causal inference to multimodal social-media data.

Currently I’m an AI Architect and Cyberinfrastructure Facilitator at GSU’s ARCTIC HPC cluster (~75M CPU and ~1.2M GPU hours annually). I’m Co-PI on a $984k NSF CyberTraining proposal on humane AI agent design, a Steering Committee member of the IEEE working group on AI for Human Flourishing, and selected for OSG School 2026 and the NAIRR-supported AI Unlocked workshop.

Explore my Research, my Resume, and my Blog.

News