Building AI systems for uncertain environments, from UAV autonomy research to enterprise agentic workflows.
Beyond the résumé: who is actually doing the work.
Computer Science & Engineering graduate from the University of Toledo. My work spans autonomous drone systems, enterprise AI, sports analytics, and data engineering. Not by design, but by instinct: I go where the hardest problems are.
From a PostgreSQL security database with zero data loss across every disaster-recovery test, to Oracle Cloud enterprise applications at First Solar - the through-line is engineering for the moment things go wrong, not just the demo. Human-in-the-loop by design; recovery mechanisms as first-class features.
A system someone can't override isn't autonomous - it's unpredictable. The recovery mechanism is part of the design, not an afterthought.
Real edge cases, latency requirements, stakeholders who need to understand the output. That pressure makes the work honest.
A drone that navigates unseen worlds, a city reporting tool in production downtown, an insurance model that watches itself drift, a chemical car that won worldwide. Not by design, but by instinct: going where the hardest problems are - and shipping.
The skill that transfers between fields isn't domain knowledge - it's knowing how to ask the right question when you don't have home-field advantage.
An ACM Computing Surveys taxonomy of how AI actually fails. Field notes on what "agentic" really means in production. Quantum computing certificates earned for curiosity, not the résumé. The questions compound faster than the answers - that is the point.
The skill that transfers between fields isn't domain knowledge - it's knowing how to ask the right question when you don't have home-field advantage.
A system someone can't override isn't autonomous - it's unpredictable. The recovery mechanism is part of the design, not an afterthought.
The stack as a constellation: clusters that orbit each other, with stars sized by how often they carry real work.
Career and education as one continuous line. Newest first, then back to the origin story.
Each project a different operating condition. Hover a poster for the story; click for the full picture.
Published work and ongoing investigations.
Thirty-seven real AI failures, distilled into the four ways autonomous systems actually break.
Training distributions that diverge from deployment conditions. The initiating mechanism in 18 of 37 documented failures.
Monitoring blind spots that allow small deviations to compound undetected until failure becomes irreversible.
Systems that perform on benchmarks but fail at the boundary of their training manifold under novel real-world inputs.
Emergent failure modes from agent interactions producing unsafe equilibria that are absent in single-agent testing.
Certifications, awards, and the communities behind the work.
Thinking in public: dispatches from the edge of AI, robotics, and real-world systems.
Press, recognition, and - as they arrive - words from the people behind the work.
Endorsements from collaborators land here as they arrive.
Not tutorials, but perspective. The mechanisms behind the work are sketched in code and traced back to the project that taught them.
What I'm actively learning, building, and chasing outside of work.
Whether it's a full-time opportunity, research collaboration, or just a conversation about autonomous systems, I'm listening.
