Founding Machine Learning Engineer

Wholesail•San Francisco, CA

20h

About The Position

Wholesail is building a financial network from the ground up that connects the systems of vendors and buyers involved in wholesale trade to enable streamlined payment and the transfer of risk to third parties. This will allow vendors to offload risk and eliminate tens of billions of waste — while giving creditworthy buyers better terms and third-party capital, unlocking hundreds of billions (ultimately trillions) in additional sales. The primitives of this network scale across industries and geographies: a universal approach to ERP integrations, modern payment rails, and a live trade-credit bureau to underwrite risk — which we're calling Lighthouse. The Opportunity: Risk & Capital Products. Credit is the load-bearing beam of our network. Every time a vendor ships goods before getting paid, someone is taking a risk — today it's the vendor, tomorrow it should be a third party at a fair price. Getting that transfer right is what unlocks the next order of magnitude of sales across the wholesale economy, and the only way to get it right is to underwrite buyers more accurately than anyone else in the industry. Through Lighthouse, we're building a live, reciprocal trade-credit bureau: vendors on our network contribute real-time payment behavior on a long tail of SMB buyers that no traditional bureau sees. That data — combined with the bank, ERP, and transaction signals already flowing through Wholesail — is a modeling dataset that doesn't exist anywhere else. The first MLE on this team gets to decide what we build with it. The problems are real and the stakes are significant. Our models directly shape the terms buyers are offered and the losses Wholesail and our capital partners absorb. There's no established playbook here and no legacy stack to inherit — you'll be setting the direction for how we do modeling, data engineering, and production ML at Wholesail for years to come.

Requirements

5+ years of experience building models for production use cases.
Deep modeling skill: strong command of supervised learning on tabular data, including feature engineering, model selection, hyperparameter tuning, calibration, and rigorous offline and online evaluation.
Strong data engineering skills — comfortable owning ETL and feature pipelines end-to-end against real, messy production data (SQL and a modern data-processing stack).
Proficiency in Python and the modern ML ecosystem (pandas, scikit-learn, PyTorch or TensorFlow, XGBoost/LightGBM, Jupyter, etc.).
Solid statistical reasoning — you can spot leakage, selection bias, label noise, and spurious correlations, and you know why your offline metric doesn't always predict online performance.
Track record of owning models end-to-end: not just prototyping in a notebook, but taking something from a problem statement to a production system that makes decisions.
Excellent written and spoken English communication skills; ability to explain model behavior, tradeoffs, and limitations to non-ML audiences (product, ops, capital partners, leadership).
The spirit of a team player who believes in fostering a healthy and supportive work environment.
BA or BS in Computer Science, Statistics, Mathematics, a related technical field, or equivalent practical experience.

Nice To Haves

Direct experience in credit risk modeling — PD/LGD/EAD models, scorecards, underwriting models, exposure management, or collections modeling — particularly for SMB or subprime segments.
Experience in fintech, lending, payments, fraud, or insurance, or with regulated modeling environments (model risk management, adverse action, fair lending).
Experience building or operating ML platform components — feature stores, model registries, training orchestration, online serving, monitoring/drift detection.
Backend engineering depth: comfortable writing production services, designing APIs, and reasoning about distributed systems that consume your models.
Strong data analysis background — fluency in exploratory analysis, experimentation, and translating data into product and business decisions.
Experience with LLMs and agentic systems applied to operational problems (document understanding, KYB, entity resolution, agent tooling).
Experience as the first or founding ML hire at a company, or early on a team that had to build ML infrastructure from scratch.
Published research, Kaggle placements, or prominent OSS contributions in the ML or data ecosystem.
Advanced degree (MS or PhD) in a quantitative field.

Responsibilities

Own credit risk modeling end-to-end — from problem framing and data pipeline design, through model development and validation, through production serving and monitoring.
Design, build, and validate credit risk models (PD, LGD, EAD, fraud, exposure sizing, pricing) against our proprietary reciprocal-bureau data and external signals.
Choose the right tool for the job — gradient-boosted trees, deep models, LLMs, or classical statistical models — and defend the choice.
Build the pipelines, feature logic, and training datasets that your models (and future models) run on.
Ship models into production systems where they make real decisions on real dollars.
Own deployment, monitoring, drift detection, and iteration.
Partner closely with product, engineering, capital markets, and external lending and insurance partners.
Explain your models to people who care about the outcomes but not the math, and translate their constraints back into modeling decisions.
Help define the hiring bar and interview process for the data scientists and MLEs who will join after you.
Have a strong voice in what this team becomes.