Lead Data Scientist

Department: Data Science

Location: San Francisco, New York

Compensation: $210K – $250K • Offers Equity

Employment Type: FullTime

About Middesk:

Middesk makes it easier for businesses to work together. Since 2018, we’ve been transforming business identity verification, replacing slow, manual processes with seamless access to complete, up-to-date data. Our platform helps companies across industries confidently verify business identities, onboard customers faster, and reduce risk at every stage of the customer lifecycle.

Middesk came out of Y Combinator, is backed by Sequoia Capital and Accel Partners, and was recently named to Forbes Fintech 50 List.

About The Role:

We are actively building AI-driven applications that streamline customer workflows, focusing on business onboarding. With our proprietary identity data assets and deep domain expertise, we are uniquely positioned to expand into a broader set of AI-powered solutions that drive long-term growth.

We’re looking for a hands-on applied ML expert to help build the technical foundation for these efforts. Ideally you have shipped external-facing models in the risk/fraud space and know the messy realities of imbalanced data, low labels, and changing behavior. This is a highly technical, hands-on role with wide influence on how we design, build, and scale ML at Middesk.

We follow a hybrid work model, and for this role, there is an expectation of 2 days per week in our SF/NYC office. Candidates should be based within a commutable distance, as we believe in the value of in-person collaboration and building strong team connections while also supporting flexibility where possible.

What You'll Do:

• Build risk & fraud ML applications: Deliver production ML models in fraud, trust & safety, KYB, and compliance domains, with measurable impact on customer workflows.

• Tackle hard data problems: Work on classification problems with extreme class imbalance, sparse signals, and “cold start” label challenges.

• Innovate in feature engineering & labeling: Use graph-based techniques, weak supervision, LLMs, and AI agents to improve signal extraction and automate labeling process.

• Establish ML infrastructure foundations: Partner with the ML infra team to design feature services, model training pipeline, model serving standards, and orchestration to scale multiple ML use cases.

• Design and implement knowledge graph solutions: Leveraging LLMs for graph construction, querying, and retrieval to enhance entity resolution and business identity use cases.

What We're Looking For:

• 7+ years of production ML experience in one or more of the following areas:

  • Building Production ML for risk, fraud, credit, or trust & safety: Track record of shipping external-facing ML applications in one or more of these domains.

  • Knowledge graph applications: Hands-on experience building, querying, or extracting signals from knowledge graphs—ideally over business entity networks (companies, persons, addresses, relationships) to support identity verification, fraud detection, or risk decisioning.

  • Entity resolution for business or individual identities: Experience disambiguating and linking records across noisy, incomplete, or conflicting data sources—particularly in KYB, KYC, AML, or identity verification contexts where the same real-world entity may appear under different names, addresses, or tax IDs.

• Expertise in classification with real-world ML challenges, for example: imbalanced labels, sparse signals, cold start, and production version management.

• Hands-on ML infrastructure experience: feature stores, model management, ML training/serving pipelines.

• Comfort as a senior IC: setting technical direction, mentoring peers, and establishing best practices.
  

Nice-To Have:

• B2B SaaS experience, ideally building ML products for enterprise customers.

• ML pipeline and automation engineering: Experience building end-to-end training harnesses that automate feature engineering, data validation, and model training.

• Experience scaling ML across multiple products or risk domains.