I help industrial SMEs and deep-tech hardware startups build robust embedded systems.
Fifteen years of safety-critical embedded work, focused now on the verification loop that lets AI coding agents build, test, and iterate on real embedded C++, without a human in the middle of every cycle.
What this practice is
Two things, and they reinforce each other: deep safety-critical embedded experience, and the infrastructure that lets AI coding agents work on that kind of code without breaking it. If your embedded codebase has outgrown ad hoc development, whether legacy debt, reliability problems, or a team that wants to move faster, that's the work I take on.
Safety-critical embedded
Fifteen years of C++ in systems where a defect that reaches the field is expensive to fix and hard to live with.
Agent-ready verification
Building the build-and-test feedback loop that makes AI coding agents genuinely usable on embedded codebases.
Measured by what ships
Audit and development engagements with acceptance criteria per milestone: progress you can see, not hours logged.
Latest posts
- Constexpr Testing Finds Compiler and Standard Library Bugs Without a Board
Constexpr tests run inside the compiler, so cross-compiling them finds toolchain bugs with no hardware. Three real cases from a CBOR decoder: an arm-gcc std::string constant-evaluation failure, a -Wuseless-cast conflict across word sizes, and std::isnan not being constexpr.
- Rolling Out Coding Agents to an Engineering Team, Champions First, Then Batches
A coding agent rollout works best champion-first: one engineer per team proves the agent on a real workflow, then onboards peers in small batches.
- Compile-Time Testing in C++: Prove It, Don't Just Test It
Compile-time testing with consteval and static_assert, proven against a real CBOR (RFC 8949) codec and a 20-plus-compiler CI matrix that caught real constexpr-evaluator bugs in shipping toolchains before they shipped.