Senior Applied Data Scientist

Belvo

About The Position

Belvo is an open finance API platform working to democratize access to financial services in Latin America. The Senior Applied Data Scientist will join the Intelligence Squad, which builds AI capabilities and develops models using financial data and payment information aggregated across the region. The role focuses on applying data science to consequential financial decisions, including credit risk, fraud detection, transaction categorization, and income verification. The successful candidate will serve as the squad’s data science reference, deciding which problems merit modeling, defining standards for model quality, and raising data science practice across collaborators. This is a hands-on role spanning problem framing, statistical and machine-learning model selection, feature engineering, experimentation, and production ownership. The position works closely with Product, Engineering, Data Research, the Intelligence Squad, business stakeholders, and sometimes customers. It requires translating ambiguous business needs into modeling plans and communicating trade-offs and impact to technical and non-technical audiences. The role also expects daily use of agentic coding and research tools and continued engagement with developments in generative AI. Belvo uses Python for data science and modeling, SQL for data access and feature work, AWS managed services, an experimentation platform, and Datadog for observability. Teams follow continuous integration and continuous delivery practices. Belvo is remote-friendly, has offices in São Paulo and Mexico City, and aligns across time zones for at least three hours per day.

Requirements

At least 5 years of experience as a Data Scientist, Applied Scientist, or in a similar hands-on modeling role.
Experience owning models end to end in production, from problem framing through live monitoring, with measurable business impact.
Strong grounding in statistics and ability to explain how models work at a statistical level.
Experience building and comparing multiple model types for the same problem, such as tree-based methods, regression, and deep learning, and justifying the selected approach.
Proficiency in Python for data science, including pandas and scikit-learn, and experience with at least one of TensorFlow or PyTorch.
Proficiency in SQL for extracting and shaping data.
Experience personally taking several models into production.
Understanding of engineering fundamentals, including testing, version control, CI/CD, and maintainable code.
Experience building LLM applications beyond basic API calls, including tool use or agents, retrieval, and structured outputs, with at least one application shipped to production.
Experience creating an evaluation approach to assess whether an LLM application is good enough.
Daily use of agentic coding tools, with the ability to explain a concrete workflow improvement and how outputs are checked.
Proficiency in written and spoken English.

Nice To Haves

Experience in fintech, financial risk, or another regulated data environment.
A business or strategy background alongside data science, such as an MBA, consulting experience, or a product or strategy role.
Experience in a customer-facing data science role, such as forward-deployed engineering or an applied/solutions team at a data platform company.
Experience with MLOps tools for model versioning, monitoring, and automated retraining.
Experience deploying models on AWS, such as with SageMaker or an equivalent managed ML service.
Experience writing evaluations or an LLM-as-judge harness for a non-deterministic system, and informed views on its limitations.
Experience using LLMs in data pipelines for labeling, extraction, or categorization, including understanding where they outperform classical models and where they do not.

Responsibilities

Lead the design and selection of statistical and machine-learning models, including classification, regression, ranking, NLP, and LLM-based approaches, for use cases such as risk scoring, fraud detection, and transaction categorization.
Use agentic coding and research tools to explore data, test feature hypotheses, scaffold pipelines, and build services around models.
Partner with Product and business teams to identify meaningful signals in financial data and translate ambiguous questions into well-scoped modeling problems.
Build or shape pipelines, APIs, and services that run models reliably in production, and own live models through monitoring, drift management, retraining, and retirement.
Design and run experiments, including A/B tests and offline evaluations, to determine whether model changes improve intended metrics.
Work with Engineering, Data Research, Product, the Intelligence Squad, and business stakeholders to deliver end-to-end solutions grounded in business needs.
Engage stakeholders and customers when needed, and clearly explain modeling trade-offs and impact to non-technical audiences, including executives.
Keep current with generative AI model releases, research, and tooling, and share useful developments through demonstrations, documentation, or company-wide AI Chapter sessions.
Decide which problems are worth modeling; recommend rules, heuristics, or better data pipelines when they are more appropriate, and retire models when evidence supports doing so.

Benefits

Stock options.
Annual company bonus linked to company performance.
Flexible working hours.
Remote-friendly work arrangement.
Pet-friendly workplace.
Health insurance.
Paid time off on your birthday.
Laptop renewal every three years.
Training budget.
Team-building events.
Option to swap bank holidays within 30 days.
Fitness and wellness stipends.
Yearly company offsite.
Fresh fruit and all-you-can-drink tea and coffee.
Extra days off accrued after completing two, three, four, and subsequent years at Belvo.
Yearly department offsite.