PhD Research Intern – Graph Learning & Agentic AI for Fraud Detection

Socure is building the identity trust infrastructure for the digital economy — verifying 100% of good identities in real time and stopping fraud before it starts. The mission is big, the problems are complex, and the impact is felt by businesses, governments, and millions of people every day.

We hire people who want that level of responsibility. People who move fast, think critically, act like owners, and care deeply about solving customer problems with precision. If you want predictability or narrow scope, this won’t be your place. If you want to help build the future of identity with a team that holds a high bar for itself — keep reading.

Socure is seeking a PhD Research Intern to join our Fraud Data Science team for Summer 2026. This internship offers the opportunity to work on frontier machine learning research at the intersection of graph-based learning (including GNNs and graph transformers) and agentic AI systems, applied to large-scale, adversarial fraud detection problems.

Fraud detection presents uniquely challenging research conditions: dynamic and heterogeneous graphs, extreme class imbalance, evolving adversaries, weak supervision, and real-world deployment constraints. We are looking for a researcher who is excited to tackle these challenges and push the state of the art in graph representation learning and autonomous AI systems.

The goal of the internship is to develop novel modeling approaches that can lead to both patent filings and academic publications, while influencing next-generation fraud detection systems at production scale.

This role is ideal for a PhD candidate who wants to combine deep technical rigor with real-world impact in a high-stakes domain.

• Formulate and drive original research directions in graph learning for fraud detection, exploring architectures such as GNNs, graph transformers, and hybrid models

• Design scalable approaches for dynamic, heterogeneous, and large-scale fraud graphs

• Investigate agentic AI and LLM-augmented systems for automated risk reasoning, investigation workflows, and decision support

• Develop robust learning techniques for adversarial and non-stationary environments

• Conduct rigorous empirical evaluation on real-world, production-scale datasets

• Translate research insights into practical system-level implications in collaboration with data scientists and engineers

• Contribute to patent development and preparation of submissions to top-tier academic venues

Minimum Qualifications

• Currently pursuing a PhD in Computer Science, Machine Learning, Statistics, Mathematics, or a related field

• Strong research foundation in one or more of:

  • Graph representation learning (e.g., GNNs, graph transformers)

  • Transformer architectures and deep learning

  • LLMs and agentic AI systems

  • Adversarial, robust, or trustworthy machine learning

• Demonstrated research capability (e.g., publications, preprints, or equivalent work)

• Strong programming skills in Python

• Experience with modern ML frameworks (e.g., PyTorch, TensorFlow, JAX)

• Ability to independently scope and execute open-ended research problems

Preferred Qualifications

• Publications at top-tier ML/AI/data mining conferences (e.g., NeurIPS, ICML, ICLR, KDD, WWW, WSDM, ACL)

• Experience scaling graph-based or transformer-based architectures to large datasets

• Familiarity with graph learning libraries (e.g., PyG, DGL)

• Experience working with noisy, highly imbalanced, or adversarial datasets

Location: US Remote

Duration: 12–16 weeks (Summer 2026)

Socure is an equal opportunity employer that values diversity in all its forms within our company. We do not discriminate based on race, religion, color, national origin, gender, sexual orientation, age, marital status, veteran status, or disability status.
If you need an accommodation during any stage of the application or hiring process—including interview or onboarding support—please reach out to your Socure recruiting partner directly.

Follow Us!

YouTube | LinkedIn | X (Twitter) | Facebook