Описание

Belvo is an open finance API platform that democratizes access to financial services in Latin America. The Intelligence Squad is Belvo's AI enabler team, with this role focusing on building models from financial data and payments to make real decisions about creditworthiness, fraud detection, and transaction categorization.

Responsibilities

Lead the design of statistical and ML models (classification, regression, ranking, NLP or LLM-based) for problems like risk scoring, fraud detection and transaction categorization, choosing approaches based on merit and defending choices with data.
Use agentic coding and research tooling as a core part of the work, including exploring data, generating and testing feature hypotheses, scaffolding pipelines, and writing services around models.
Work with Product and business teams to identify signals in Belvo's financial data that explain outcomes and translate ambiguous business questions into well-scoped modeling problems.
Build or shape the pipelines, APIs and services that run models reliably in production, and own them: monitoring, drift, retraining, and retirement.
Design and run experiments (A/B tests and offline evaluation) to confirm that model changes improve metrics.
Work closely with Engineering, Data Research, Product, and the rest of the Intelligence Squad to ship end to end solutions.
Work directly with stakeholders and customers, translating ambiguous business questions into well-scoped problems and explaining trade-offs and impact to non-technical audiences.
Stay current with GenAI developments, including model releases, papers, and tooling, bringing useful insights back to the squad.
Decide what is worth modeling, pushing back when rules, heuristics, or better data pipelines can solve a problem without a model, and retiring models when evidence dictates.

Requirements

5+ years of experience as a Data Scientist, Applied Scientist, or in a similar hands-on modeling role.
Owned models end to end in production, from problem framing to live monitoring, with measurable business impact.
Solid grounding in statistics and ability to explain how models work at that level.
Built and compared multiple model types (e.g., tree-based methods, regression, deep learning) for the same problem and justified the chosen approach.
Comfortable in Python for data science (pandas, scikit-learn, and at least one of TensorFlow or PyTorch) and SQL for data manipulation.
Taken several models into production independently, understanding engineering fundamentals: testing, version control, CI/CD, and maintainable code.
Built with LLMs beyond API-call level (tool use or agents, retrieval, structured outputs) and shipped at least one to production, with established evaluation methods.
Agentic coding tools are part of daily workflow.
Proficient in written and spoken English.

Nice to have

Experience in fintech, financial risk, or another regulated data environment.
Business or strategy background alongside data science (MBA, consulting, or product/strategy role).
Worked in a customer-facing data science role (e.g., forward-deployed engineer or solutions/applied team).
Experience with MLOps tooling for model versioning, monitoring, and automated retraining.
Deployed models on AWS (e.g., SageMaker or equivalent managed ML service).
Written evaluations or an LLM-as-judge harness for non-deterministic systems, with opinions on their limitations.
Used LLMs inside the data pipeline itself (labeling, extraction, categorization), understanding their strengths and weaknesses compared to classical models.

We offer

Stock options.
Annual company bonus linked to company performance.
Flexible working hours.
Remote friendly.
Pet friendly.
Health Insurance.
Paid time off on your birthday.
Laptop renewal every 3 years.
Training Budget.
Team building events.
Bank holidays swap within 30 days.
Fitness/wellness stipends.
Yearly company offsite.
Fresh fruit every week, all-you-can-drink tea and coffee.
Extra days off accrued when completing 2, 3, 4 years and beyond at Belvo.
Yearly department offsite.
Python as the primary language for data science and modeling (pandas, scikit-learn, and deep learning frameworks).
SQL against a cloud data warehouse for data access and feature work.
Claude deployed company-wide and Claude Code for every engineer.
Cloud infrastructure on AWS, favoring managed services.
An experimentation platform for A/B testing and offline evaluation.
Datadog for service observability.
Continuous Integration and Continuous Delivery best practices.
Primarily use Python on the backend with Django and Python's asyncio.
Javascript, Vue.js and Sass on the frontend, leveraging a custom design system.

Hiring process

People team chat.
Take-home challenge.
Challenge presentation.
Team fit chat.
Meet the founders.