Machine Learning Engineer

Human time and effort is the rarest thing in the universe. The fact that so many people hate what they do for work is a disaster on so many levels. How many people live quiet lives of misery? How many generational talents do we miss due to circumstance? Dex's mission is to close the gap between Talent & Opportunity. We believe that work, and being productive, is part of the human condition, that regardless of AI, human passion, ingenuity and a desire to do things means that work will always exist.

We are building Dex to create a future where technology helps every individual to understand their strengths, interests and unique abilities, and to connect with opportunities where they can thrive, because more people doing things they love leads to a better, happier and more productive world. We're starting by connecting the world's most ambitious software engineers with the companies that deserve them.

About Dex: Dex is backed by a16z Speedrun, Concept Ventures and angels from OpenAI, Wise, ElevenLabs, Meta's board, and many more. We're an ambitious, direct and kind group of people, we value authenticity, ambition and people who make things happen. We're a team of 12 (and growing) based in London (Borough). We work together 3 times a week because we value in-person time, but we're person-centric and flexible. We hire adults who can manage their own schedules.

About the Role: This is an ML engineering role, not a research role, not a prompt engineering role. You understand the underlying mechanics: how embeddings encode meaning, how attention shapes retrieval, how to build and evaluate representations that power matching and ranking systems. You use LLMs where they're the right tool, but your first instinct is to understand the problem at a model level, not to reach for an API. The work is concrete: building a system that takes a new role and stack-ranks the entire candidate database against it in seconds. You own the representations and scoring models that sit at the core of how Dex connects engineers with companies. The matchmaking engine you build here is the foundation that powers everything that comes next — candidate-facing products, automated outreach, smarter sourcing — so you're building for durability, not just the immediate use case. Your ML judgment is backed by solid engineering execution.

What You'll Do:

• Own the matchmaking engine — build and improve the AI-driven talent matching system; design representations, scoring models, and instrumentation from the ground up.
• Work with embeddings and retrieval — build and evaluate embedding models, vector search, and semantic retrieval systems that power candidate-to-role matching.
• Design and run evaluations — build practical eval frameworks for model behaviour and output quality; make rollout safety and failure handling first-class concerns.
• Contribute to agent and LLM systems — work on the voice agent backend and LLM pipelines with a model-level understanding of what's actually happening.
• Ship reliable backend services — build production-grade Python services, not prototypes; handle errors, retries, latency, and observability as standard.

About you:

• ML foundations — you understand how embeddings, attention, and retrieval systems work at a model level; you can reason about representations, not just API responses.
• Production judgment — you know when to use an LLM, when to use classical ML, and when to use neither; you've made these calls in production, not just in notebooks.
• Evaluation & guardrails mindset — you build evals before things go wrong; you design failure handling and rollout safety into systems from the start.
• Backend execution strength — you ship reliable, maintainable services; your track record shows production ML systems, not just impressive demos.
• Strong technical depth in: Python backend (FastAPI or equivalent, async patterns, Postgres, Redis), Embeddings, vector search, and semantic retrieval (building and evaluating, not just calling), LLM integration with model-level understanding (attention, context windows, trade-offs).

Nice to Have:

• Experience with recommendation systems, ranking models, or candidate/item matching.
• Classical ML background (supervised/unsupervised, feature engineering, gradient boosting).
• Experience with voice agents or real-time audio pipelines.
• Familiarity with Pydantic AI or similar agent frameworks.
• Experience in recruiting tech or marketplace systems.

What we offer:

• Above market salary — we're building a world-class team and believe pay should match that ambition.
• Significant equity — we want this to be the last job you 'have' to take.
• Full private healthcare & dental — because life happens and we want to support you.
• Fertility benefits & enhanced parental support.
• Gym membership & wellness benefits.
• Claude tokens — lots and lots of Claude tokens.
• Regular socials, offsites and an annual trip.
• MacBook Pro & team kit.
• 'Whatever you need' approach to tools and support.