Machine Learning Evaluation Engineer in London

AI Evaluation, Research Methods, Python, LLM Observability

Salary range: £60,000-£80,000 p.a. + equity, depending on experience (up to £100,000 for candidates with exceptional relevant experience)

What is Marker? Marker is an AI-native Word Processor – a reimagining of Google Docs and Microsoft Word. Join us in building the next generation of agentic AI assistants supporting serious writers in their work. We are a small, ambitious company using cutting-edge technology to give everybody writing superpowers.

What you'll do at Marker:

• Design and implement evaluation frameworks for complex, subjective AI outputs (like writing feedback that's meant to inspire rather than just correct).
• Build flexible evaluation pipelines that can assess quality across multiple dimensions - from human preference to actual writing improvement.
• Research and prototype new evaluation methodologies for creative and subjective AI tasks.
• Collaborate with our engineering team to integrate evaluation insights into our development process.
• Help define what "quality" means for different AI outputs and create metrics that actually matter for our users.
• Work on challenging problems like: "How do we automatically evaluate whether an AI comment successfully encourages thoughtful revision?"

What we can offer:

• A calm, human-friendly work environment among kind and experienced professionals.
• Fun, creative, novel, and interesting technical work at the intersection of AI research and product development.
• An opportunity to work with and learn about the latest advancements in AI evaluation and language models.
• Direct collaboration with leadership to shape how we understand and improve our AI systems.
• As much responsibility and growth opportunities as you want to take on.

Are you a good fit for this role? In order to be successful in this role, you will recognise yourself in the following:

• You have experience with AI/ML evaluation methodologies and can speak the language of AI research.
• You've worked hands-on with language models and understand the challenges of evaluating subjective, creative outputs.
• You are a self-starter willing to work independently and at speed - we imagine a 2-week experiment cadence at most.
• You are familiar with and have worked on related technical systems (evaluation pipelines, data collection tools) but don't need to be a full-stack engineer.
• You think critically about what metrics actually matter and aren't satisfied with vanity metrics.
• You're comfortable working with ambiguous problems where the "right answer" isn't obvious.
• You have some programming experience (Python preferred) and can work independently on technical projects.
• You're interested in the intersection of AI capabilities and human creativity.

An exceptional candidate for this role would be able to demonstrate some of the following:

• Experience building evaluation systems for generative AI in production environments.
• Knowledge of TypeScript and ability to integrate with our existing systems.
• Background in human-computer interaction, computational creativity, or writing research.
• Experience with A/B testing, statistical analysis, and experimental design.
• Familiarity with modern AI observability and monitoring tools.
• Published research or deep interest in AI evaluation methodologies.
• Interest in writing (fiction, non-fiction, essays).

However, you are NOT expected to:

• Be a senior software engineer - we're looking for someone who can build evaluation systems, not architect our entire backend.
• Have solved every evaluation problem before - this is cutting-edge work and we're figuring it out together.
• Be experienced with every library in our stack from day one - you'll work closely with Ryan and our engineering team.
• Have a specific degree - we value practical experience and research ability over credentials.

Our stack:

• Our AI engine uses a range of models, including self-hosted and fine-tuned open source models, as well as latest reasoning models from Anthropic and OpenAI.
• Evaluation and research tools built primarily in Python, with integration into our TypeScript infrastructure.
• Our agentic AI execution platform is written in TypeScript, hosted on Cloudflare Workers.
• Standard ML tooling: various evaluation frameworks, data analysis tools, and monitoring systems.
• Our text editor frontend is a web application built with React, TypeScript and ProseMirror.

Apply now! Interested? Email us at work@writewithmarker.com with your CV (or a link to your CV site). Tell us a little bit about yourself and why you'd like to work at Marker!

Please note that this role is currently only available based in our London hub, and at this time we are not able to sponsor work visas in the UK.