Learning Scientist

career; Mountain View (CA); On-site; USD 140,000 - 200,000; Full time

Job summary

LearnVector in Mountain View, CA is hiring for a role focused on designing AI-native teaching methods and building an evidence-based measurement backbone. You will run studies with real learners, craft reliable assessments, and ensure product decisions are grounded in data and literature.

The role emphasizes rigorous research, collaboration with engineers, and the dissemination of findings to guide product strategy and learning outcomes.

Qualifications

• Deep grounding in learning science and evidence-based practice.
• Strong experimental design and statistics skills.
• Experience with human subjects research.
• Ability to work with data (Python or R) and with engineers.
• Excellent communication of evidence to non-specialists.

Responsibilities

• Invent and prototype AI-native teaching methods with engineers.
• Design the measurement backbone for learning outcomes.
• Run studies with real learners across pilot to longitudinal cohorts.
• Build assessments with validity and reliability as engineering requirements.
• Set the evidence bar with literature and data; be open to being wrong.
• Collaborate with the founding team on product decisions based on evidence.

Skills

Learning science

Experimental design

Human subjects research

Python or R

Communication

Education

PhD in learning sciences or related field

Tools

Python

R

Job description

About LearnVector

For most of history, great teaching has been scarce. A brilliant teacher who knows you well, adapts to how you learn, and patiently stays with you until you get there — almost no one has had that. AI changes what's possible. LearnVector, founded by Andrew Ng, is building a trustworthy AI guide for learning, with a mission to accelerate human development. We're a small, fast-moving team working on-site in Mountain View, California and backed by a $100 million investment from Coursera.

About the role

You will invent new ways to teach that take advantage of agentic AI — and apply rigorous measurement to prove they work. Agentic AI makes teaching moves possible that no classroom or MOOC could offer: a tutor that remembers everything, infinitely patient practice, feedback on real work product, assessment woven invisibly into learning. Most of these possibilities are unexplored, and much of what's shipping across the industry today has no evidence behind it.

Your job is both halves: design the new methods, and hold them to the standard of evidence. What did the learner retain a week later? Can they apply it to work that looks nothing like the exercise? You'll be the person in the company whose answer to "is this teaching?" is a measurement, not an opinion.

What you will do

• Invent and prototype AI-native teaching methods — working with engineers to build them into the product, not writing papers about what could be built
• Design the company's measurement backbone: what we measure to know learning happened (skill gain, retention, transfer), and how it's instrumented into the product
• Run studies with real learners — from one-week pilots to longitudinal cohorts — sized and designed so results mean something; kill designs the evidence doesn't support, including your own
• Build assessments worth trusting: performance tasks and rubrics that measure real competence, with validity and reliability treated as engineering requirements
• Set the evidence bar company-wide: when the team debates a pedagogical choice, you bring the literature and the data, and you're open to being wrong
• Work directly with the founding team, including Andrew, on what we build and what we believe; your evidence shapes decisions at the top, not just recommendations that get filed

What you bring

• Deep grounding in learning science — the experimental literature on how people acquire and retain skills (retrieval, spacing, feedback, transfer, expertise development) and where its limits are
• Strong experimental-design and statistical skills: you know what a well-powered study needs, and you notice when a result is noise dressed as signal
• Research experience with human subjects — lab or field — and the pragmatism to run informative studies inside a fast-moving product, not just ideal ones
• Enough technical fluency to work with data directly (Python or R) and to collaborate closely with engineers on instrumentation
• Excellent communication: you make evidence legible and actionable to a non-specialist team

Nice to haves

• PhD in learning sciences, cognitive psychology, education, or a related field — or equivalent research experience
• Experience with intelligent tutoring systems, adaptive learning, or AI-based instruction
• Psychometrics and assessment-validity experience (IRT, rubric calibration, rater reliability)
• Experience measuring learning in adult professional or workplace contexts, where completion and retention behave nothing like the classroom

What success looks like

In your first 30 days, you will have defined the first version of our learning‑outcome measures and have a study running with real learners.

In 6 months, the company will make product decisions against evidence you produced, at least one novel AI-native teaching method you designed will be live in the product, and we'll know — with data — whether it teaches better than what it replace you…

Equal opportunity

LearnVector is committed to a workplace of mutual respect and equal opportunity. We hire based on qualifications, merit, and business needs, and do not discriminate on the basis of any characteristic protected by applicable law.

Accommodations

If you need a reasonable accommodation at any point in the application or interview process, we'll work with you. Requests are kept confidential and separate from hiring decisions.