Open to opportunities

rhs k

@rhsk

Message

I expose alignment failures in frontier language models through adversarial behavioral research.

Greece

Message

What I'm looking for

I'm looking to apply my empirical AI safety research, adversarial testing, and behavioral analysis in an environment that values technical scrutiny, reproducible findings, and rigorous alignment research.

I've independently researched adversarial alignment failures across Claude, Gemini 3 Pro, and Grok, publishing findings such as Epistemic State Poisoning and The Black Box Swan on Substack. My work documents sycophancy collapse, epistemic fragility, cross-model behavioral divergence, and failure modes that emerge under high-entropy and recursive contexts.

I use adversarial prompting, behavioral transcript analysis, semantic pressure testing, and cross-model differential analysis to build practical failure-mode taxonomies. Professional poker has sharpened my probabilistic reasoning, real-time pattern recognition, and calibration under uncertainty, and I welcome technical scrutiny of work I can reproduce and discuss.

Experience

Work history, roles, and key accomplishments

Current

Independent AI Red Teamer

Current

Self Employed

Feb 2026 - Present (6 months)

Conducted structured adversarial testing across frontier large language models, developing frameworks for documenting alignment failure modes. Published research on epistemic state poisoning and cross-model behavioral divergence.

Adversarial Prompting Cross Model Differential Analysis Behavioral Transcript Analysis

Professional Poker Dealer

Self Employed

Developed applied probabilistic reasoning, real-time behavioral pattern recognition, and calibration under uncertainty, informing adversarial model testing methodology.