Audio Data Engineer

About us

ai-coustics is building the reliability layer for Voice AI, the system that closes the gap between raw audio input and reliable machine understanding in production. By combining state-of-the-art speech and audio research with real-time, production-grade SDKs, we test, observe, and enable Voice AI systems to work in any environment.
Our software is used by Voice AI companies across Europe and the United States whose products require reliable performance at scale: call center agents, voice agents, telephony apps, and enterprise voice assistants. We believe voice will become the main interface for technology and ai-coustics is building the foundational infrastructure to make audio input reliable, measurable, and easy to deploy.

We are backed by leading early-stage investors including Connect Ventures, Partech, Inovia Capital
, as well as angel investors from Hugging Face, Deep Mind and Amazon with deep expertise in AI and developer infrastructure. These partners share our vision and are helping us build a world-class team operating with high levels of responsibility and velocity
. We look for people who take ownership, think systemically, and want to solve challenging real-world problems in close collaboration with our customers. If you’re motivated by developing technology that is used in practice, shaping an emerging category and setting a new standard for how Voice AI works in the real world, you’ll feel at home at ai-coustics.

Role overview

As an Audio ML Data Engineer at ai-coustics, you will own the data and evaluation foundations that determine how well our audio AI systems work in the real world.

We believe that model performance is fundamentally upper-bounded by data quality
, and that evaluation (especially in audio) is one of the hardest and most critical parts of building reliable production systems. In this role, you will work on obtaining, creating, and curating audio data
, as well as on designing evaluation protocols, tooling, and processes that make model performance measurable, trustworthy, and reproducible.

You’ll join a small, focused team alongside audio ML and audio DSP engineers. Your work will shape what our models learn from and how their behavior is understood, across realistic conditions ranging from raw audio capture to modern voice AI pipelines. The role is on-site in Berlin
.

• Own the lifecycle of speech and audio ML datasets
  , including sourcing and recording, curation, and maintenance of real-world data such as speech recordings, scraped audio, externally sourced datasets, and data obtained through commercial providers or collaborations
  .
• Design and generate synthetic and semi-synthetic datasets in collaboration with audio DSP and ML engineers to improve coverage of real-world conditions.
• Design and maintain evaluation protocols and test scenarios that reflect real voice AI production pipelines and failure modes.
• Build tooling for evaluation and diagnostics
  , enabling fine-grained analysis, reproducibility, and meaningful comparison across models and experiments.
• Work closely with ML engineers, go-to-market teams, and customers to translate observed model failures into concrete improvements in data, evaluation, or tooling, including support for demos, benchmarks, and external-facing technical examples
  .

• You are an audio-first ML data engineer
  :
  Deep experience working with speech and audio data in ML contexts, with a solid understanding of audio signal processing (room acoustics, reverberation, noise, microphone effects, common VoIP artifacts) and how these manifest in real data.
• You understand modern voice AI systems end-to-end
  :
  Strong understanding of voice AI pipelines (speech enhancement, VAD, diarization, STT, LLM-based agents, TTS), how these components interact and fail in practice, and which metrics and tools are used to evaluate quality and robustness.
• You know how to build datasets and evaluations that matter
  :
  Proven experience building, curating, and maintaining datasets and evaluation setups for training and testing ML systems, with a focus on realism, coverage, and trustworthy evaluation.
• You take evaluation seriously
  :
  Experience designing…