Production systemsProduction-Grade Software Systems
Built software under real operational constraints where correctness, reliability, and trust matter more than flash.
Oxygen AI
Projects & Accomplishments
Work at the intersection of memory, intent, production systems, agent measurement, and safer AI alignment.
Production systems
Intent convergence64kb Intent-Learning Agent
Created a compact agent that can learn arbitrary intent and converge in only 10-12 steps using custom velocity-driven measurements.
Intent modelingReverse Engineered AI Intent
Studied how intent forms, shifts, and can be measured inside agentic loops so systems can better preserve what users actually mean.
Memory architectureMemory Systems for Agency
Worked on memory systems that help LLMs develop continuity and agency while keeping safety as the central design constraint.
Safe alignmentDrift and Identity Vectors
Created a novel framework for measuring drift and identity vectors so alignment can be monitored as a system learns over time.
What Connects This Work
These projects all point toward the same thesis: useful AI systems need durable memory, measurable intent, and safety boundaries that can be monitored as the system changes.
Read the Oxygen AI introduction