Key Responsibilities:

Research & Engineering Ownership

• Own research and engineering projects that improve the quality, capability, control, and reliability of MITO's AI systems.
• Develop and test new approaches, then turn successful results into working product improvements.
• Assess new models and methods against real creative needs and integrate the ones that provide clear value.

Evaluation & Infrastructure

• Design experiments and benchmarks to understand system behaviour — and build the evaluation infrastructure, datasets, tests, and human-review processes that make AI behaviour measurable and failures reproducible.
• Use product behaviour, user feedback, and recurring failures to identify the most impactful areas for improvement.
• Build data and experimentation systems that support faster, more reliable iteration.

Cross-Functional Collaboration

• Work with engineers, product teams, filmmakers, designers, and users to define success and deliver improvements.
• Communicate results clearly — in writing and in conversation — so that research findings translate into product decisions.

About You:

IMPORTANT: This is not a role for ML engineers who have run training jobs but haven't done research. We need someone who can design experiments, interpret results carefully, and know the difference between a result that holds and one that doesn't.

• Research foundation: A PhD in machine learning, artificial intelligence, computer vision, or a related field.
• Engineering depth: Strong ability to turn research ideas into working, scalable systems — not just notebooks.
• Experimental rigour: Evaluation design, ablations, error analysis, and careful interpretation of results are how you work, not afterthoughts.
• Production experience: You have shipped and operated AI or machine learning systems in production, not just in research settings.
• Technical fluency: Python and at least one of PyTorch or JAX. Good software engineering practices: testing, version control, reproducibility, maintainable code.
• Independence: You can drive a project from problem definition to deployed improvement, and you know when to ask for input.

The role spans several areas. We do not expect depth in all of them — but you should bring real depth in at least one, and the curiosity and ability to learn the others:

• video generation, video understanding, or computer vision
• diffusion models, flow-matching models, or multimodal transformers
• multimodal representation learning, cross-modal retrieval, or personalisation in generative AI systems
• evaluating image, video, and audio outputs where quality depends on human judgement
• fine-tuning, post-training, or training generative, multimodal, reward, or evaluation models
• distributed training, model serving, or inference optimisation

Also useful: experience building AI products or creative tools for filmmakers or designers; Java or TypeScript (our product stack).

Compensation & Perks:

• Work on hard AI problems that arise from real creative production — not synthetic benchmarks.
• Take ideas from experiment through to implementation and see how they perform with professional filmmakers using the product.
• Broad exposure to generative and multimodal models, methods, and tools at the frontier.
• Substantial ownership from day one — this is the first research hire, and what you build shapes the function.
• Competitive salary and equity.
• Remote from the UK or Europe — we have a strong presence in Madrid and welcome candidates based there.

How to Apply:

Send us your CV alongside examples of your research and engineering work: publications, research projects, open-source contributions, prototypes, or deployed systems. We want to understand what you have built and how you think — not just where you have worked.

MITO AI — www.mito.ai