Job details
About this role
Role overview Join an early-career research team building foundation models for financial data, where several models already run in production on bureau, transactional, and graph datasets. The work extends these models to new domains and downstream tasks such as credit risk, customer acquisition, anti-fraud, and LTV forecasting. This is a hands-on research position with full ownership from idea through production impact.
Responsibilities - Train self-supervised models on discrete sequences targeting state-of-the-art performance on fintech downstream tasks - Track current research and apply modern NLP and deep learning techniques to financial modeling problems - Work with large multimodal datasets spanning tabular, behavioral, transactional, device, network, text, time series, and graph data - Optimize compute utilization for both training and inference workloads - Own projects end-to-end, including data collection, experimentation, training, inference optimization, and impact evaluation - Conduct rigorous evaluations and contribute to written research outputs and conference talks
Requirements - Degree in mathematics, engineering, computer science, artificial intelligence, or another strong quantitative field - Solid foundation in machine learning and deep learning, with hands-on experience building and experimenting with PyTorch and LLM-based models - Understanding of dataset design and data mixing for deep learning training - Demonstrated ability to maximize compute utilization and run experiments efficiently - Strong communication skills and a high degree of autonomy and ownership
Nice to have - Published research in machine learning or deep learning - Personal projects or startup experience in AI, ML, or DL - Internship experience on research, ML, or DL teams - English proficiency at B1 or higher for collaborating across an international team
Benefits and work setup - Relocation support to hubs in Serbia, Georgia, Mexico, or Colombia, with assistance for the employee and family - Flexible work from offices or remotely within GMT-2 to GMT+5 time zones - Healthcare coverage, an education budget for language lessons, training, and certifications, and a wellness budget for mental health and fitness - 20 days of annual leave plus paid sick leave