Easy JobPrincipal Research Scientist [Risk]
Banco Plata, S.A., Institución de Banca Múltiple · Worldwide · Full-time · Remote
About Plata
Plata is one of the fastest-growing fintech companies in the world. In just 3 years, we've grown to 3M+ customers and reached a $5B+ valuation. We're now strengthening our Risk & Decisioning core team and are looking for a Principal AI Engineer to set the technical bar and build best-in-class, production-grade models that materially move business metrics.
Why this role
• You will own our foundation model program end to end - from research direction to models running in production and driving real credit decisions.
• Your models directly move the numbers that matter: cost of risk, approval rates, and portfolio NPV across a fast-scaling credit portfolio.
• You'll work with one of the richest financial datasets in LATAM billions of transactions and behavioral events across 3M+ customers, growing daily.
Challenges that await you
• Build an end-to-end foundation model over financial event sequences - transactions, credit bureau data, and in-app behavioral events with subsequent fine-tuning for downstream business tasks: underwriting (PD), credit limit strategy, fraud detection, collections, and propensity models.
• Drive technical decisions end-to-end: methodology → implementation → performance and latency → robustness, interpretability, and regulatory compliance.
• Take models from research to production: training infrastructure, evaluation frameworks, model serving, latency/cost optimization, and monitoring.
• Research state-of-the-art approaches in the industry, publish your own work, and speak at leading conferences.
• Mentor senior engineers and scientists; own technical standards for model development across the team (design reviews, evaluation methodology, deployment practices).
• Communicate results clearly to cross-functional stakeholders: product, risk, business, and leadership.
What makes you a great fit
• Proven experience applying deep learning to sequential data - transformer architectures on event/transaction seque