Cloud platforms
AWS, Kubernetes and Terraform systems designed around blast radius, operability and cost—not a pristine diagram.
Cloud platforms, data systems and AI agents—built with failure modes, security and the bill in view.
01 / SYSTEMS I BUILD
AWS, Kubernetes and Terraform systems designed around blast radius, operability and cost—not a pristine diagram.
High-volume ingestion, search and retrieval architectures that stay useful when the data gets large, late and weird.
Agents, RAG and tool-driven workflows with explicit boundaries, durable execution and evidence you can actually inspect.
Low-latency media and operational systems where protocols, payloads and production constraints all matter at once.
02 / SELECTED WORK
A mix of personal systems, corporate engineering and field-tested architecture. Specific enough to be credible; discreet where the work demands it.
A Rust-native agent runtime built for work that outlives a chat turn.
Queued research and media jobs, cancellation semantics, artifacts, provenance and operational APIs—currently being hardened in closed alpha ahead of a future public release.
Internal engineering for multi-agent development workflows.
Ongoing corporate work around execution boundaries, release discipline and operator safety—deliberately light on public detail while the system is still taking shape.
A realistic physics-based game about detecting incoming rockets and putting an interceptor on the right trajectory.
A realtime browser simulation pairing believable flight physics with streamed video, telemetry and per-frame metadata—all engineered under strict latency constraints.
One local-first operating layer for every machine in the workshop.
A concept for coordinating printers, lasers, inventory, maintenance, job history and documentation without turning the workshop into somebody else’s cloud dependency.
03 / OPERATING DOCTRINE
The job is not to make complexity disappear. It is to put it somewhere deliberate, observable and survivable.
Name them. Test them. Revisit them before they become folklore.
Design the degraded path while everyone is still calm.
A system nobody can afford to run is just an expensive drawing.
Identity, authorization, data residency and blast radius belong in the first conversation.
04 / TRAJECTORY
Cloud, data and AI systems · Client-facing technical strategy · Hands-on delivery
Teams, migrations and platform delivery · Turning ambiguity into operating systems
Three years owning technology direction, delivery and the consequences
End-to-end product engineering · Frontend, backend, data and the infrastructure underneath
Datacenters, networks and virtualization · Debugging from first principles
05 / THE PHYSICAL LAYER
Fabrication is the shortest possible feedback loop: model, manufacture, measure, fail, revise. It keeps the systems thinking honest—and occasionally produces something useful.
Two multi-material workhorses, thousands of print hours, engineering polymers and the occasional machine rebuild.
Cut, engrave, instrument, repair and improve. The physical layer is where tolerances stop being theoretical.
A local-first control plane for machines, materials, jobs, maintenance and workshop history.
Architecture, technical leadership, agent systems, or a platform that has started growing teeth.