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 The AI Engineer will build the agentic systems at the core of the product: systems that understand each learner, plan a path with them toward skills worth having, and work with them step by step until they get there. This is not a wrap-an-API role. The hard problems are the ones frontier models don't solve on their own: maintaining an accurate picture of a learner over weeks and months, deciding what to teach next and when to hold back, keeping long-running conversations useful rather than merely pleasant, and verifying that generated teaching is correct before a learner ever sees it. You'll own systems end to end — design, implementation, evaluation, and iteration against real learner data. What you will do - Design and build the agentic core: multi-step tutoring loops, tool use, memory, and planning over long-horizon learner relationships - Build the learner model — the evolving, evidence-backed representation of what each learner knows, wants, and responds to — and the systems that read and write it - Build evaluation harnesses for conversational quality and teaching quality, and use them to drive iteration; define what "this session taught something" means operationally and measure it - Design guardrails and verification layers so generated content and tutor claims meet a bar a trusted brand requires - Work daily with the founding team, including Andrew, on the hardest product questions: what should an AI tutor do, and how do we know it's working? What you bring - AI-native: you default to AI-assisted coding and building agentic automations in everything you do, you have an appetite for and record of experimenting with the newest AI engineering practices - Experience as a software engineer, with substantial hands-on work building with LLM APIs (Claude, OpenAI, or similar): agentic workflows, tool use, structured output, long-context and memory patterns - Experience shipping and operating LLM systems in production, including evaluating them - you have opinions about evals because you've built them - Strong Python and/or TypeScript/Node engineering skills; comfort owning services end to end - Ability to turn a fuzzy product question ("is the tutor actually helping?") into a measurable system, and ship without heavy oversight Nice to haves - Experience with conversational AI products, tutoring systems, or long-running assistant relationships - Background in recommendation, personalization, or user-modeling systems - Familiarity with the education or learning-science landscape - Experience with voice interfaces or real-time interaction What success looks like In your first 30 days, you will have shipped a measurable improvement to the core tutoring loop and stood up an evaluation that tells us whether it worked. In your first 6 months, the agentic core — learner model, planning, verification — will be a durable system the whole product builds on, with quality metrics the team trusts and a cadence of improvement driven by real learner data. 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.

Requirements

Must-haves

The AI Engineer will build the agentic systems at the core of the product: systems that understand each learner, plan a path with them toward skills worth having, and work with them step by step until they get there.
Design and build the agentic core: multi-step tutoring loops, tool use, memory, and planning over long-horizon learner relationships
Build the learner model — the evolving, evidence-backed representation of what each learner knows, wants, and responds to — and the systems that read and write it
Build evaluation harnesses for conversational quality and teaching quality, and use them to drive iteration; define what "this session taught something" means operationally and measure it
Design guardrails and verification layers so generated content and tutor claims meet a bar a trusted brand requires
Work daily with the founding team, including Andrew, on the hardest product questions: what should an AI tutor do, and how do we know it's working?
AI-native: you default to AI-assisted coding and building agentic automations in everything you do, you have an appetite for and record of experimenting with the newest AI engineering practices
Experience as a software engineer, with substantial hands-on work building with LLM APIs (Claude, OpenAI, or similar): agentic workflows, tool use, structured output, long-context and memory patterns
Experience shipping and operating LLM systems in production, including evaluating them - you have opinions about evals because you've built them
Strong Python and/or TypeScript/Node engineering skills; comfort owning services end to end
Ability to turn a fuzzy product question ("is the tutor actually helping?") into a measurable system, and ship without heavy oversight

Nice-to-haves

Experience with conversational AI products, tutoring systems, or long-running assistant relationships
Background in recommendation, personalization, or user-modeling systems
Familiarity with the education or learning-science landscape
Experience with voice interfaces or real-time interaction
What success looks like
In your first 30 days, you will have shipped a measurable improvement to the core tutoring loop and stood up an evaluation that tells us whether it worked.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aiapillmpythontypescript

Platform: lever
Effort: quick