Senior Machine Learning Engineer, Model Risk Management
Block, Inc. is a technology company building tools for economic empowerment, operating a portfolio of financial brands including Square, Cash App, Afterpay, TIDAL, Bitkey, and Proto. It serves sellers, consumers, artists, and bitcoin users through payments, banking, buy-now-pay-later, music streaming, and bitcoin self-custody and mining products.
About Block, Inc.
Block, Inc. builds technology aimed at increasing access to the global economy. Its brands each unlock different aspects of the economy: Square makes commerce and financial services accessible to sellers; Cash App is an easy way to spend, send, and store money; Afterpay (Clearpay in the UK) helps customers manage spending over time; TIDAL is a music platform empowering artists as entrepreneurs; Bitkey is a self-custody bitcoin wallet; and Proto builds open, accessible bitcoin mining hardware and services. Block has been a long-time advocate for bitcoin, integrating bitcoin buying/selling and Lightning Network payments into Cash App and Square, funding open-source bitcoin infrastructure through Spiral (including the Lightning Development Kit), and engaging in bitcoin policy advocacy through groups like COPA, the Digital Energy Council, the Crypto Council for Innovation, and the Texas Blockchain Council. The company also runs a significant open source program (contributing to projects like Goose, an on-machine AI developer agent, OkHttp, Retrofit, gRPC, Envoy, and MySQL) and an AI research effort, including co-founding the Agentic AI Foundation with Anthropic and OpenAI. Block serves individual consumers, small and medium-sized businesses/merchants, artists, and developers.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will independently challenge model owners across lending, fraud, and AML; reproduce their results, set and defend the acceptance thresholds, and own the call on whether a model is sound. Hunt silent errors that make metrics lie, evaluate models under real-world conditions, ship production validation tooling, build agentic validation systems, assess ML systems end to end, connect explainability and fair-lending findings to product decisions, and help define standards for validating production AI.
Requirements
- Quantitative degree or equivalent experience and senior individual-contributor depth building or validating models in credit, fraud, or financial crime.
- Experience with effective challenge methodology, reproduction, conceptual soundness review, benchmarking, stress testing, and outcomes analysis.
- Deep applied machine learning and statistics across regression, tree ensembles, and deep learning.
- Strong experimentation and statistical rigor, including holdout design, uncertainty, calibration, and generalization.
- Production-quality Python, SQL on large datasets, reproducible code, and testing practices.
- Fluency building with LLMs and agentic tools, with judgment about trustworthy outputs.
- Familiarity with model risk management frameworks and fair-lending standards.
- Strong communication skills and independence operating under ambiguity.
Responsibilities
- Independently challenge model owners across lending, fraud, and AML; reproduce results, set acceptance thresholds, and decide whether models are sound.
- Identify silent errors that mislead metrics and prove them out before production.
- Design evaluation for rare events, shifting populations, and post-launch drift.
- Work hands-on in unfamiliar codebases and ship production validation tooling.
- Build agentic validation tooling that orchestrates parallel agents.
- Evaluate ML systems end to end across features, training, serving, monitoring, and scale.
- Connect explainability and fair-lending findings on consumer credit models to model and product decisions.
- Help define standards for validating production AI systems.
Benefits
- Remote work
- Medical insurance
- Flexible time off
- Retirement savings plans
- Modern family planning
