Senior ML Data Scientist (Credit Risk)

Zed•San Francisco, CA

About The Position

Zed is building the first AI-native, licensed neobank in the Philippines designed to democratize access to premium financial services for young professionals in global markets. The current banking system is broken, often shutting out the world’s youngest and fastest-growing consumer classes—we’re here to fix it. Our team is uniquely positioned to solve this. We are Stanford engineers and former YC founders who have spent our careers at the intersection of banking and hyper-growth startups like Square, Facebook, and Box. We’ve been here before, having previously built and exited Symple (YC W'17), a fast-growing B2B payments company. We are backed by world-class investors, including Accel, Valar, Immad Akhund (Mercury), Dalton Caldwell (Y Combinator), and Kunal Shah (Cred). The Role Zed underwrites credit using foundation models that profile risk from transaction data, financial documents, and other structured and unstructured sources — not credit scores. That means our ML stack looks less like a traditional bank's and more like a modern AI system: embedding models, transformer architectures, and LLM-assisted data pipelines sitting alongside classical credit and fraud models. We're hiring a Senior ML Data Scientist to work directly with our data lead across all of it — core credit models, account management, and fraud detection — with a particular focus on pushing the frontier of how we represent and reason about financial data. This is a senior role, which means we expect you to have opinions about the stack, shape how we build, and set the technical bar for ML at Zed as the team grows. If you're the kind of person who's deploying neural networks and transformer-based models in production rather than just reading about them, this role was written for you.

Requirements

7+ years of experience in applied ML or data science with a focus on credit risk, fraud, or financial services
Hands-on experience with LLMs, embeddings models, or transformer-based architectures — not just familiarity, but production or near-production deployment
Proficiency in Python and SQL; experience with frameworks such as XGBoost, LightGBM, PyTorch, or similar
Strong feature engineering skills — you know how to extract signal from messy, sparse, or heterogeneous financial data
Solid statistical foundation: anomaly detection, supervised classification, model calibration, experimentation design
Experience building and monitoring production models including alerting on performance degradation and concept drift
Familiarity with both rule-based and model-driven approaches — and when to use each
A point of view on how ML systems should be built — you can articulate tradeoffs, push back on bad decisions, and bring junior team members along
Comfort operating as a senior ML voice at an early-stage company — you own problems end-to-end and set the standard for others

Nice To Haves

Experience with emerging markets or data-sparse environments is a plus

Responsibilities

Work closely with our data lead on the full risk model suite: core credit decisioning, account management, and fraud detection
Own data preparation pipelines for model inputs — including using NNs and LLMs to represent transaction data as vector embeddings for quantitative analysis
Experiment with and deploy neural network and transformer-based architectures in the underwriting process
Build agent scaffolding and harnesses within underwriting workflows — this is active, in-production experimentation, not research
Design and deploy fraud detection models combining rule-based systems and ML to identify suspicious activity in real time
Engineer features from structured and unstructured data sources
Develop monitoring systems to keep models accurate and reliable in production
Partner with engineering and risk operations to integrate model outputs into decisioning systems
Influence technical direction — you'll have a seat at the table when we make decisions about how ML is built and deployed at Zed