About us

Encord is the universal data layer for AI that helps 300+ AI teams train and run models on the right data. Our platform indexes, curates, annotates, and evaluates data across the full AI lifecycle, from development through production.

Trusted by Woven by Toyota, AXA, UiPath, Zipline, and more. We're an ambitious team of 100+ working at the frontier of AI and have raised $60M in Series C funding from Wellington Management, CRV, Next47, and Y Combinator.

The role

We're hiring a Learning & Development Specialist in London to design and deliver how Encord teaches the specialist workforce behind its human data — the annotators producing work for frontier AI labs and physical AI companies, in domains such as medical imaging and robotics, and on data types we are running for the first time. How well that workforce is taught is the largest controllable driver of the quality we ship.

You will work at the point where a technical specification becomes something a non-specialist can execute correctly: converting dense, sometimes ambiguous annotation guidelines into instruction a distributed workforce absorbs in days rather than weeks, and into assessment that proves they have.

That teaching problem is a real one, and it is why we are hiring for learning-science expertise rather than for training delivery. The people you teach are capable adults, but they are not domain experts, they come from a wide range of educational backgrounds, and they are frequently working in their second language. Managing cognitive load, breaking a specification into modules that build on each other, leading with worked examples before independent practice, and testing for what actually predicts performance on the job — that craft is the role.

It is a hands-on role. You will design the material and you will stand up and teach it, and it is measured in operational terms: ramp time to competence, first-pass quality, and how much of the workforce is qualified for the work we are selling.

What you'll do

• Turn a client specification and a set of annotation guidelines into teachable material within days of a project launch — modular, sequenced, and built on worked examples, edge-case libraries, and annotated failure cases
• Design against explicit learning-science principles rather than instinct — manage intrinsic and extraneous cognitive load, chunk a dense specification into modules that build, scaffold then fade support, and use spaced retrieval so what is taught in week one survives to week four
• Run live training sessions yourself, remotely and to distributed cohorts, daily during a project ramp — and adjust in the room when a concept is not landing
• Build criterion-referenced assessments that decide readiness rather than test recall, with pass thresholds set alongside the Quality Systems Lead — so that nobody works a project queue without having demonstrated they meet the standard
• Design for a novice audience by default: no assumed prior knowledge, plain language, and material that works for someone reading in their second language
• Build durable curricula for the domains we sell repeatedly — medical imaging, document AI, LLM evaluation and red-teaming, robot teleoperation, egocentric capture — with progression into higher-skill, higher-rate work
• Coach the Human Data Program Managers into better teachers: session formats, feedback technique, and observation against a standard, since much of the instruction is delivered by them rather than by you
• Sit with annotators and watch where they get stuck, then redesign — most of what you need to know is visible in the first hour someone spends on a new task
• Partner with Product on what belongs inside the Encord platform as guidance at the point of work rather than in a course nobody re-opens
• Measure every programme against a before-and-after in quality and throughput terms, and retire the ones that do not move an operational number