Sardine AI Labs is now accepting applications for up to five research fellowships, open to student or faculty researchers in the U.S. or Canada who are focused on some of the hardest problems in preventing fraud and financial crime.
Frontier AI labs have shown the immense benefits of having multi-modal AI models which allow you to extract intelligence across different modalities of audio, video and text. We think that the next unlock in fraud and financial crime prevention will occur from creating a multi-modal AI that goes across device, identity, behavioral, and transaction data. This is what excites us the most about the potential of our AI Lab, and we can’t wait to see what our research fellows build with us.
The program builds on Sardine AI Labs’ work in modeling financial behavior across device, identity, behavioral, and transaction data. The selected research ideas should advance AI/ML model capabilities while addressing the demands of production risk systems, including accuracy, latency, explainability, and governance.
Selected researchers will work with Sardine’s AI Labs team to build new foundation models and are highly encouraged to publish their work in peer-reviewed conferences and journals. Currently we are only looking for AI researchers from U.S. or Canadian universities. Applications may be open to more countries in the future.
Research areas
Proposals should address one or more of the following:
- Cross-institution generalization: Transfer knowledge across card issuers, including institutions unseen during training.
- Multi-stream modeling: Learn from connected sequences of transactions, device activity, identity signals, and customer behavior.
- Scaling laws: Understand how performance changes with larger models, more data, and longer histories.
- Identity matching: Improve matching across names, addresses, and other identifiers.
- Graph learning: Model payment relationships to detect money-laundering patterns.
- Real-time inference: Serve large models with long customer histories at low latency.
- Cold-start learning: Improve predictions for customers or institutions with limited history.
- Applications beyond fraud: Extend financial representations to credit risk, churn, forecasting, and related tasks.
Eligibility requirements
- Advanced-degree student, researcher, or faculty at a U.S. or Canada academic institution.
- Background in machine learning, deep learning, large language models, computer science, engineering, or a related quantitative field.
- Availability to work from Sardine’s Berkeley or Toronto office for 3-9 months as an internship for students or a sabbatical for faculty.
- For applicants in the United States:
- U.S. citizen or permanent resident, or authorized to work as an intern in the U.S.
- For applicants in Canada:
- Canadian citizen or permanent resident or authorized to work as an intern in Canada