I'm an independent security researcher focused on smart-contract auditing, exploit analysis, and adversarial AI workflows. My work centers on finding real paths to improve consumer AI systems — building benchmarks and workflows that dramatically raise the accuracy of frontier models in niche fields and precision-based tasks.

I'm especially interested in agentic security, quantization, harness-engineering, low-level protocol logic, evaluation design, and model improvements on limited compute. I also build tooling around bug bounties, agent-based workflow reviews, vulnerability-pattern extraction, and triage-grade reporting.

This blog is the logbook: notes on building AI systems on a budget, the architecture decisions behind them, and what actually works when the model stays the same but the system gets smarter.