Who we are

A European defence startup based between Paris and Warsaw, building modern software tailored for hostile environments. Our customers include European military and police units. We are in stealth but have already been recognised as one of France's top 20 seed companies. Backed by Europe's leading defence VC and angels including founders of Cloudflare, Quantum Systems, and ElevenLabs.

Our founding team includes two early Cloudflare employees and a former GIGN commander (French special forces). Other team members include ex-Googlers, Thales PhDs, and ex-founders. We do not work on weapons systems or hardware products.

The role

This is an empirical engineering role, not a research post: you'll ship. But the engineering is experimental by nature - most of your time goes to forming hypotheses, measuring them on real field data, and iterating on what the field sends back.

Your first mission: our real-time speech-to-text system (Python).

Our users aren't speaking into clean conference-call microphones. They're on military radio, in noisy environments, using dense operational jargon, on hardware that runs entirely on-premise, with limited compute. Off-the-shelf models fall apart in these conditions.

What you'll work on

Evaluate and adapt commercial and open-source speech models, and build benchmarks to measure them honestly on real field data
Own the pipeline around the model: preprocessing degraded and shifting signals, voice activity detection, transcription, diarisation, and fine-tuning models to the domain
Turn raw operational recordings into clean datasets
Ship all of it into a system running live under real operational constraints

Speech is where you start. As the product grows, so will the role.

Who we're looking for

Pragmatic, product-minded, strong execution skills, and a desire to build from zero
Owned at least one ML or data-heavy project end-to-end, from an ambiguous problem statement to measured results
Scientific mindset: you define your own evaluation metrics, design experiments, and draw reliable conclusions from noisy data
Experience implementing Voice AI systems in resource-constrained environments
Comfortable diving into audio processing and low-latency systems
Thrive in ambiguous, fast-moving environments
Want your work to have real-world impact, helping those who serve
French is a plus, not required

Location and compensation

Paris, France, on-site; relocation support available
60,000 to 80,000 EUR plus equity, depending on experience