Founding AI/ML Engineer, Authorization
Engineering, Full-time, Founding role, San Francisco, CA (onsite), $140K to $200K
About Lane
Lane is building the financial infrastructure for the agentic web. Our mission is to turn intent into action: we give AI agents a dependable way to reach businesses, to prove who they act for, and to pay for what they buy. The first web was built for eyes and fingers; it was a library of documents. The second was built for apps; it was a set of walled gardens. We're building the third, a web built for agents.
Lane has two products. The Lane Wallet connects a person's own card to their agents and lets them approve what an agent buys. The Lane Index lets agents find businesses and their tools, then authenticate to them and pay for them, with agent auth and billing built in. We're a small team in San Francisco, and the people who join now take real ownership of the systems, the customers, and the company we become.
The Role
You'll lead how Lane decides whether an agent is allowed to do what it's about to do. A person approves an intent in plain language: "book me a flight to Austin under $400." An agent then takes dozens of actions on their behalf. Your models decide, at each step, whether the action still matches what the person approved, how risky it is, and when to stop and ask.
This is a new problem. Card networks score a transaction; nobody yet scores an agent's action against a person's intent. You'll define the approach, build the first models, and own the data and evaluations that prove they work. The base salary range is $140K to $200K, plus competitive equity. We'll talk specifics during the process.
What You'll Work On
Intent to authorization
Turn a person's natural-language approval into a precise, checkable envelope: merchants, amounts, items, substitutions, and time windows. Decide what the model infers and what the person must confirm.
Action scoring
Score each agent action and each purchase against its approved intent. Catch drift, substitutions, prompt injection, and over-spend before a payment goes through.
Risk and trust signals
Combine agent identity, user history, device signals, and merchant data into risk decisions: approve, step up to the person, or refuse.
Evaluation
Build the eval sets, labeling, and offline and online metrics for authorization quality. Track false approvals and false blocks as product metrics, not research metrics.
Production ML
Ship models that run inside the purchase path under tight latency and cost budgets. Choose when a rule, a small model, or an LLM is the right tool.
What We're Looking For
- 2+ years building ML or LLM systems, in industry or research, with at least one that made real decisions in production.
- Depth in at least one of: fraud and risk modeling, policy and authorization systems, or LLM evaluation and guardrails.
- Strong applied judgment about LLMs: where they are reliable, where they fail, and how to measure it.
- Solid software engineering in Python or TypeScript. You ship your own models.
- You design evals before you design models.
- Clear writing. You can explain a model's failure modes to a compliance officer and to a customer.
Nice to Have
- Payments fraud, chargebacks, or card network risk experience.
- Work on agent safety, prompt-injection defense, or tool-use policies.
- Familiarity with authorization standards: OAuth scopes, policy engines, mandates.
How to Apply
To apply, email team@getonlane.com with your resume attached, plus a short note answering the questions below.
- Something you've built, run, or launched that you're proud of. Include a link or a short description of what it did and what you were responsible for.
- Which area above you'd most want to own, and why.
- Anything else you'd like us to know.
A resume and LinkedIn are preferred, and links to your work are welcome.
Interested?
Email your resume and a short note to team@getonlane.com. Questions about the role go to the same place.
Lane is an equal opportunity employer. We do not discriminate on the basis of race, color, religion, creed, sex, sexual orientation, gender identity or expression, pregnancy, national origin, ancestry, age, disability, medical condition, genetic information, marital status, veteran or military status, or any other status protected by law. We evaluate every candidate on the merits of their work and the contribution they would make to the team. We will consider qualified applicants with arrest and conviction records in accordance with the San Francisco Fair Chance Ordinance. If you need an accommodation at any step of the hiring process, email team@getonlane.com.