Remote EU
Note - whilst the role is remote in Europe it requires monthly travel to Germany to spend time in person as a whole team
About the company
We’re an early-stage AI company building a new class of creative software powered by large-scale generative models.
Our work focuses on transforming visual inputs into rich, high-quality audio outputs, enabling new kinds of creative workflows across media, entertainment, and interactive applications. We operate at the intersection of research and product, turning foundational model advances into systems that can be deployed reliably at scale.
The team is small, highly technical, and deeply collaborative, spanning research, engineering, and product.
The opportunity
This is a hands-on, research-driven role at the core of how our generative models are designed, trained, and evolved.
You’ll work on multimodal generative systems that connect vision, sound, and temporal structure. The role spans the full model lifecycle — from data curation and experimental design to training, evaluation, and deployment — with a strong emphasis on rigorous experimentation and practical impact.
You’ll collaborate closely with other researchers and engineers, explore new modelling ideas, and help translate research insights into production-ready capabilities.
What you’ll do
* Design, implement, and train large-scale multimodal generative models for audio-related tasks.
* Explore new modelling approaches for sound, speech, and music generation, drawing inspiration from advances in language and vision models.
* Develop and experiment with post-training techniques to unlock new capabilities such as controllability, editing, and structured generation.
* Run focused ablation studies, analyse results, and communicate clear, actionable insights to guide future research.
* Contribute across the full model development pipeline, including data curation, experimentation, evaluation, and deployment.
* Work closely with engineering partners to ensure research translates into robust, scalable systems.
You should have
* Hands-on experience training large-scale generative models in a fast-moving research or applied setting.
* Deep understanding of modern machine learning methods in at least one domain (e.g. language, vision, video, or audio).
* Strong proficiency with PyTorch, transformer-based architectures, and contemporary deep learning tooling.
* Solid understanding of distributed training techniques such as data/model parallelism and low-precision training.
* A strong track record working on generative models, demonstrated through publications, open-source contributions, or impactful applied work.
* Ability to work independently while collaborating effectively in a small team.
Tech stack & research environment
* PyTorch and modern deep learning frameworks.
* Transformer-based and diffusion-style generative models.
* Distributed training across multi-GPU systems.
* Large-scale data pipelines for training and evaluation.
* Close integration between research code and production systems.
Nice to have
* Experience profiling and optimising GPU workloads.
* Strong software engineering skills and experience contributing to large, shared codebases.
* Prior work on audio, speech, or music generation models.
* Familiarity with audio representations, codecs, or signal processing concepts.
Why it’s exciting
* Work on foundational generative models with real-world deployment.
* High ownership and autonomy from day one.
* Direct influence over research direction and technical strategy.
* Close collaboration between research and product.
* Competitive compensation and equity aligned with impact.
#J-18808-Ljbffr