Staff Machine Learning Engineer, Radar

Description

Who we are

About Stripe

Stripe is a financial infrastructure platform for businesses. Millions of companies—from the world’s largest enterprises to the most ambitious startups—use Stripe to accept payments, grow their revenue, and accelerate new business opportunities. Our mission is to increase the GDP of the internet, and we have a staggering amount of work ahead. That means you have an unprecedented opportunity to put the global economy within everyone’s reach while doing the most important work of your career.

The Radar ML team builds the machine learning systems that power Stripe’s Radar fraud and abuse product suite. Our models make real-time decisions across a significant share of internet commerce - helping businesses stop fraud, increase authorization, and grow safely. By turning economy-scale datasets into compelling ML products, the team directly improves customer outcomes and drives meaningful revenue for Stripe.

The team is growing, and expanding into multiple new areas. We are building abuse protections across the full customer life-cycle, adaptive and customizable fraud systems for enterprises, and risk intelligence for 50+ global payment methods.

Responsibilities

• You will lead the machine learning strategy for abuse prevention across the full customer lifecycle. Abuse doesn't stop at the payment. Bad actors create fake and duplicate accounts, take over legitimate ones, farm free trials and promotions, resell access they obtained below cost, use services they never intend to pay for, and file bad-faith refunds and disputes. For AI companies, every abusive account burns real compute, and the same actors keep moving from one account, product, and access path to the next.
• Stripe is uniquely positioned to solve this. Our network sees consumers across millions of businesses, so we can recognize a bad actor at signup, at login, at payment, and long after. Radar has already shipped the first building blocks: account evaluations at signup and login, customer-level abuse risk, and APIs that let businesses report their own abuse outcomes. You will turn these building blocks into a unified abuse prevention layer that understands the consumer behind every event and helps businesses act consistently across all of that consumer's accounts. This is a large growth opportunity for Radar, and a strategic priority for Stripe.
• This is a fast-moving 0→1 area, built alongside some of the world's leading AI companies, and the attacks are real today. Labels are sparse, delayed, and differ by business. Much of the customer journey is only partly visible to us. Adversaries adapt as soon as we ship. You will set the technical direction, make the hard architecture calls, and personally build the hardest pieces.
• Set the technical strategy and multi-quarter roadmap for lifecycle abuse ML, with product and engineering leadership: which abuse vectors to take on, in what order, and with what approach.
• Design a persistent, continuously updated consumer risk profile that combines signals from signups, logins, payments, product usage, refunds, and disputes. Evolve it from explainable composite scores toward model-driven risk scores for each type of abuse.
• Lead development of models for abuse vectors such as multi-accounting, account takeover, bots, free-trial and promotion abuse, unauthorized resale, non-payment, refund abuse, and friendly fraud. Use graph, sequence, and LLM-based approaches over payment, account, and product-usage data.
• Solve the label problem by designing ways to learn from sparse, delayed, merchant-specific, and weak labels, including merchant-shared feedback, LLM-assisted labeling, and network-level signals.
• Define how we measure success: evaluation frameworks, impact and ROI estimates, and operating-point guidance that help businesses trade off abuse losses against growth and conversion.
• Turn detection into action by shaping the risk controls, interventions, and recommendations businesses can turn on at each step of the customer journey.
• Stay deeply hands-on: write substantial production code, prototype quickly, and personally land the most technically difficult components, with the reliability and latency standards of systems in Stripe's critical path.
• Break ambiguous problems into parallel workstreams, unblock other engineers, and raise the team's bar through design review, code review, and mentorship.
• Partner directly with leading Stripe users, fraud experts, product, and adjacent Stripe teams to ground the work in real user pain and change course quickly based on what you learn.
• Build AI-native tooling, agents, and workflows that let the team investigate attacks and build and operate models dramatically faster.

Requirements

• You are a hands-on technical leader who has built ML systems that matter, shipped them to production, and led others in doing the same. You thrive in ambiguity. You can take an open-ended problem with messy data and incomplete labels, find the highest-leverage path, get to user impact quickly, and change course when the data says so. You make the engineers around you better. You're as comfortable talking with a customer about the economics of their abuse problem as you are debugging a feature pipeline. You're excited to use frontier AI both in the systems you build and in how you build them.
• We’re looking for someone who meets the minimum requirements to be considered for the role. If you meet these requirements, you are encouraged to apply. The preferred qualifications are a bonus, not a requirement.
• Minimum requirements
• 10+ years of experience developing, evaluating, and deploying machine learning models in a production environment.
• A track record of technically leading large, ambiguous ML initiatives from problem definition to production, with measurable business impact.
• Deep expertise in machine learning, statistics, model evaluation, and experiment design, including where labels are noisy, delayed, or scarce.
• Strong software-engineering skills, and you remain a hands-on contributor who regularly writes tested, maintainable production code.
• Experience designing and operating production ML systems at scale, including data and feature pipelines, real-time inference, deployment, and monitoring, and making architecture tradeoffs across them.
• Experience setting technical direction across multiple engineers or teams, and mentoring engineers to grow their scope.
• Ability to explain complex tradeoffs clearly to engineers, product leaders, and customers, and to build alignment across functions.
• Preferred qualifications
• Experience applying ML to fraud, abuse, integrity, trust and safety, identity, or security, where adversaries actively adapt to your defenses.
• Experience with entity resolution, graph learning, or behavioral and sequence modeling of users, accounts, or devices (for example, detecting multi-accounting, account takeover, or bots).
• Experience productionizing modern deep learning or LLM-based approaches, including using LLMs for labeling, investigation, or detection.
• Experience building 0→1 ML products and working directly with customers to define what to build.
• Experience building shared ML platforms, representations, or foundation models used across multiple products or teams.
• Experience building AI-assisted tooling or agents that measurably increased a team's velocity.