AI Infrastructure Engineer

Respeecher is a pioneer in the area of AI-generated voices. In 2019, our voices were the first to be used in a Hollywood movie, and in 2025 we gave AI superpowers to the main characters of the Oscar-winning movies The Brutalist and Emilia Pérez.
 

Our AI research team has developed a panoply of models and training and inference techniques that are expertly deployed by our delivery team, which has worked on 200+ projects in dozens of languages across almost every sector of the media and entertainment industry.

This year, we’re expanding beyond M&E, starting with the launch of our real-time text-to-speech API for voice chatbots.

About the role

We’re looking for an engineer to help design and code the systems we use to train and deploy AI models.

We care more about what you can do than your title or years of experience.

The current codebase is mostly a PyTorch-based monolith that has grown organically over years.

What you might work on

You won’t do all of these at once — part of your job will be to prioritize and sequence the work:

• Build a system to support processing and then training on large amounts of data stored in S3
• Design modular, composable model interfaces — e.g., DAG-style composition, DRY config, not requiring unused dependencies, separable training and inference code
• Support complex training schemes — e.g., dataset mixtures that change during training or training models to extract features for other models
• Ensure that we always know how each model was trained and on what data
• Support internal teams like a product partner: understand their workflows, build for them, teach them, and iterate based on feedback

You won’t train models, deploy them, or run inference — but you’ll build the systems that make all of that easy, reliable, and consistent across environments, with minimal opportunity for human error.

What we’re looking for

We don’t expect you to have experience with all of our tools and techniques — what matters is how you think, how you build, and what kinds of problems you want to take ownership of. You might be a good fit if:

• You have strong software engineering skills and value simplicity and clarity in code, configuration, and documentation
• You have experience training deep learning models in PyTorch and debugging performance issues with them. Ideally, you have experience with distributed training
• You understand the performance characteristics of object storage systems like S3
• You have experience with Ray
• You have experience with workflow management systems like Temporal
• You think about invariants and making systems safe by default
• You would enjoy turning complex, fragile workflows into something clean, modular, and reproducible
• You’re eager to learn what you need to do the job well — whether that’s infrastructure tools, ML workflows, or how our teams operate
• You work well in low-structure environments and like figuring out what’s most important to do
• You have strong opinions, loosely held
• You communicate clearly in English

Why you should join us

• Solve interesting problems
• Design and build substantial systems that can have a big impact on the company
• Be part of a thoughtful, low-ego team that values pragmatism, simplicity, and clarity

Benefits

• Unlimited vacation policy
• Office in central Kyiv and freedom to work from the office or remotely
• Amazing colleagues who love what they do and helping each other out!