Using AI agents safely for the development of Ada or Rust safety-critical software

AdaCore: Helping Developers Build Software that Matters
Everything we do at AdaCore is centered around helping developers build safe, secure and reliable software.

For 30 years, we've partnered with global leaders in aerospace & defense, air traffic management, space, railway and financial services. We've developed tools and services simplifying high-integrity software development through a subscription-based model. As demand for secure applications grows in industries like automotive, medical, energy, and IoT, we're adapting our proven technologies to assist a new generation of developers.
Our +190 global experts based in the US, France, Germany, the UK, and Estonia, collectively develop cutting-edge technologies to address the challenges of high-grade software development.
Joining AdaCore is about joining a culture of innovation, openness, collaboration and dependability, which defines how we work together, with our customers and partners.

Context

AdaCore makes the tools and runtime software that customers use to build software for systems where a failure can hurt people: aircraft, cars, trains. In these fields the finished system is certified against strict standards (such as DO-178C in aviation and ISO 26262 in cars), and every tool that helps build or check the software has to be trusted to a defined degree.

AdaCore's current certification framework grew up around Ada, the language long established in these safety-critical domains and the foundation of AdaCore's tooling. Newer languages are now entering the same work - most notably Rust, with its strong safety guarantees - while C and C++ remain widespread. A framework that only fits Ada is no longer enough; it has to serve Ada and Rust alike.

AI assistants (the kind built on large language models) are now good enough to help across this work: writing and reviewing code, generating tests, explaining failures. But they also make mistakes that ordinary tools do not - they can sound confident while being wrong, give different answers to the same question, and lose track of details in long documents. Recent research on AI in safety engineering makes one point especially clear: an AI model on its own has no safety net, so it must be wrapped in a system that checks what goes in, controls how it is used, verifies what comes out, and keeps a record of it all. Since engineers will use these tools with or without permission, the real question is not whether to allow them, but how to use them with enough confidence.

This internship tackles two connected questions: how to let AI assistants help with critical software while keeping the trust the standards require, and how to make the certification framework work across languages, from the established Ada to the newer Rust.

Goals

• Map the opportunities and the risks. List where an AI assistant could help across the software life-cycle, and for each, what could go wrong and how much it would matter.
• Write practical guidance. Define the checks that must surround an AI assistant - validating its inputs, controlling its use, verifying its outputs, and recording everything - so its contribution can be trusted and audited.
• Roll it out in stages. Start with the AI only advising while a human always verifies, then widen its role step by step as confidence grows, keeping the proven way of working alongside it.
• Build one concrete safeguard. A check, built into AdaCore's build system, stops any AI-generated code or test from reaching a deliverable until a human has formally reviewed it.
• Take the framework from Ada to Rust. AdaCore's certification framework was built around Ada; see how well it carries over to Rust, and to C and C++, and suggest where it needs to adapt to become genuinely language-agnostic.

• Deliver. The guidance, the rollout plan, and a small working demonstrator.

   

Skills required or nice to have

• Curiosity about software safety and certification - no prior background needed
• Interest in AI assistants and how to use them responsibly (nice to have)
• Programming in Python
• Some exposure to Ada, Rust, C or C++ (nice to have)
• Comfortable with GitLab and containers (Docker/Podman)
• Clear writing

Timeframe & Location
During 2027 - 6 months - Paris office

Beyond the job

We're a global organization driven by diverse backgrounds, fostering innovation through an open exchange of ideas. We welcome applicants of all backgrounds, celebrating diversity in ethnicity, nationality, gender, age, religion, abilities, sexual orientation, veteran or marital status. 
Our commitment is to help our teammates, wherever they are based, feel comfortable and satisfied, by encouraging flexibility to ensure them a healthy work-life balance. Additionally, we prioritize individual development by offering continuous training from day one with a personalized onboarding plan.

For more information about our recruitment process, visit our FAQ.