Built for work that can’t fail
Skeptic Dog Software is the independent practice of a software engineer who designs, builds, and ships production systems end to end. I work where the stakes are real: regulated, high-stakes, and operationally complex environments, taken from first spec to production with end-to-end ownership.
“Every engineering choice I make is a business decision — capacity sized to the shape of demand, margin designed into the architecture, and the leverage to let one engineer ship what usually takes a team.”
Failure planned for, not hoped against
I build the way fault-intolerant work demands: systems designed to hold under load, with failure planned for rather than hoped against. It’s the same instinct that draws me to compliance-critical and high-stakes domains, where the cost of being wrong is not abstract — and it shows up in the discipline behind every merge, not just the finished result.
The experience I bring
I don’t name past clients; the pattern is what matters. The engagements behind this experience share the same shape — regulated or high-stakes, operationally complex, and owned end to end — in whatever language and runtime the problem called for.
- FedRAMP-High cybersecurity — high-throughput services across DSPM, CNAPP, and a unified AI surface, under continuous-authorization constraints.
- Healthcare imaging and security — DICOM/RIS harmonization and endpoint incident-response used by hospital teams during active threat events.
- Enterprise industrial IoT — the first capability bridging cloud services with legacy on-prem for secure, multisite device discovery.
- Data and analytics at scale — multi-source ingestion and validation pipelines taken to enterprise scale, serving thousands of engineering organizations.
- Language assessment at scale — ML-driven automated assessment delivered at national exam volumes, with the ingestion and validation to match.
- Recent systems, built end to end — a white-label e-signature service, an investor data room, and an equity cap-table engine, each a Rust-serverless architecture chosen for the problem, not a house language.
The rooms this experience came from
Places I have been employed — cybersecurity and industrial control, medical imaging and data platforms. The engagements are described by domain on the work page; this is simply where they happened.
One engineer, large leverage
I run an AI-native practice by deliberate design. Development is spec-first — the feature specified and its edges interrogated before code — with fleets of agents implementing in isolated worktrees and adversarial, multi-round review gating every merge. Model output is never taken on trust: deterministic verification checks every generated artifact before it counts.
It runs on the product side too, and that half is easy to skip. Discovery starts with the person who will actually use the thing — including the non-technical operator whose workflow decides whether it gets adopted — and the same agent-driven approach that builds my software has been shipped as customer-facing product: multi-agent pipelines handling intake, drafting, editorial review, and publication end to end. A spec written without that contact is a guess with better formatting.
Two convictions sit underneath the whole practice — soak everything, because peak load arrives with peak customer stakes; and margin is designed, not trimmed, with unit costs metered as a first-class part of the pipeline and expensive compute spent only where it buys something the customer can sell. I document the model openly in an ongoing essay series on AI-native company building, talent pipelines, and adversarial agent architectures.
Based in Colorado, working wherever the problem is
Skeptic Dog Software, Ltd. Co. is based in Durango, Colorado, and works with clients wherever the hard problems live. I take on a small number of engagements at a time and hold the bar high on every one — regulated, high-stakes, and operationally complex systems, shipped end to end.