DO‑178C Test Automation
eInfochips announced NomAIzo, an AI tool that automates DO‑178C test‑case generation and claims about 50% productivity gains with 70% accuracy in safety‑critical development. The technology suggests a growing vendor push to use generative techniques to reduce manual verification effort, while leaving open questions about evidence quality and tool qualification. (x.com)
Airplane software gets certified with a paper trail that is almost as important as the code itself. In the aviation standard called DO-178C, engineers have to show that each requirement was checked, each test was run, and each result can be traced back to a specific function in the software. (rtca.org) That is why test-case writing eats so much time in avionics programs. A test case is a step-by-step recipe for proving one requirement works, like writing the exam key before the exam, and DO-178C pushes teams to build those recipes with enough rigor for certification review. (faa.gov) The hard part is not only running tests but proving you tested the right thing. DO-178C ties verification to requirement-based testing, which means the test has to come from the requirement itself rather than from a developer’s guess about what the code probably does. (faa.gov) Now vendors are trying to hand that drafting work to generative systems. eInfochips says its new NomAIzo testing tool uses agentic artificial intelligence, retrieval-augmented generation, and large language models to read requirements documents and generate test cases automatically. (einfochips.com) eInfochips says the tool is built around three software agents with separate jobs. One agent analyzes requirements, one writes the test cases, and one verifies the output before it goes back to engineers. (einfochips.com) The sales pitch is speed. In a recent company demo tied to its NomAIzo launch, eInfochips said the approach delivered about 50% productivity improvement with about 70% accuracy in safety-critical software workflows, which means people still have to review and fix a meaningful share of the output. (einfochips.com) That last number is the hinge in the whole story. In consumer software, 70% accuracy can be a useful draft; in flight software, a wrong test can create false confidence, which is worse than having no automation at all because it still produces paperwork that looks complete. (faa.gov) There is also a second certification problem hiding behind the first one. If a tool is used to eliminate, reduce, or automate verification work that humans would otherwise perform, aviation programs may have to qualify that tool under DO-330, the companion guidance for software tool qualification. (ldra.com) That means the tool itself can become part of the evidence chain. A company is not just asking whether an artificial intelligence system can write a test case faster; it is asking whether auditors will trust the generated artifact, the traceability behind it, and the process used to validate the tool that produced it. (afuzion.com) So the near-term use case is not “press button, certify aircraft.” It is “let the model draft the first version, then make engineers review, correct, and justify every line,” which is why eInfochips frames NomAIzo as a way to reduce manual effort and accelerate verification rather than replace certification engineers. (einfochips.com) What eInfochips is selling is bigger than one product demo. It is a sign that aerospace suppliers now see requirement parsing, test generation, and verification paperwork as the next place to deploy generative tools, even in one of the most regulated software environments on earth. (engineering.com)
