A clear problem. Room to explore the solution.

I enjoy investigating constraints, testing alternatives, and making the useful result dependable.

What I’m working on now →

I like the part of engineering where a concrete problem opens up several possible solutions: work out the constraint, build something small enough to test, and use the result to decide what comes next. My recent projects explore GPU scheduling, model evaluation, and reliable media processing on personal hardware.

I previously worked as a Software Development Engineer at Amazon, building and operating customer-order remediation services for Amazon Fresh. I also built and operated a community platform’s backend alongside a frontend partner for about two years. Those services supported real users; the current AI platform is independent R&D, developed alongside part-time automotive-operations work.

My background includes a Computer Science degree from Griffith University, two paid security research internships studying formal verification of smart contracts, and a fintech penetration-testing engagement in Hong Kong.

How I work: I prefer an agreed problem and clear priorities, with room to investigate the design and test alternatives. I enjoy explaining tradeoffs and making prototypes dependable enough for someone else to use.

What I am looking for: software engineering, research-software, or technical implementation work in a team that values investigation and practical delivery. AI is one area I enjoy, not a requirement: the quality of the problem and the working relationship matter more than the label.

From problem definition to operation.

I keep the original problem and success criteria visible as work moves through data, models, interfaces, deployment, and maintenance.

AmbiguityArchitecture

Turn an under-specified need into a boundary small enough to build and specific enough to test.

PrototypeOperation

Carry the awkward details—leases, recovery, provenance, failure modes—that demos can ignore.

ResultRevision

Design measurement that can disagree with the story, then change the system when it does.

Other systems I’ve built

Other work includes Email triage (a read-only nightly digest with per-verdict and per-day cost provenance; kept as a small operational note until real corrections support a quality estimate) · Universal Critic (three-pass dialectic critique: critique → steelman → adjudicate) · Researcher (search-synthesis with enforced source attribution). Smart-home and solar-sufficiency experiments remain preliminary: the former’s event bus needs repair before the hardware phase, and the latter has no reading history yet.