Senior Specialist, Data Scientist

BNY Mellon•Pittsburgh, PA

About The Position

The Payments Risk Services team is seeking a Data Scientist / Data Engineer to learn the payments data landscape and build machine learning models that detect and prevent fraudulent payments. The role is located in Pittsburgh, Pennsylvania, and is suited to a recent college graduate interested in applying data science to a high-impact, real-world problem. The role involves exploring transaction, customer, and channel data across wire, ACH, and real-time payment rails; developing supervised and unsupervised fraud-detection models; and mapping fraud typologies to detection signals and modeling strategies. The position also designs and maintains feature pipelines, works with data engineering to take prototypes into production, and explores modern AI approaches to improve detection and reduce false positives. The Data Scientist applies explainability methods so analysts, model risk teams, and regulators can understand model decisions; measures model performance and communicates impact to technical and business stakeholders; and partners with fraud subject-matter experts, product, and engineering to take projects from hypothesis through deployment and monitoring. The role also follows evolving fraud patterns, machine learning and AI techniques, and responsible-AI practices while developing skills across model science, feature science, and insight science.

Requirements

Bachelor's degree in a STEM field such as Computer Science, Data Science, Statistics, Mathematics, or Engineering, or equivalent experience.
Foundational knowledge of machine learning and statistics.
Hands-on Python coding experience with libraries such as scikit-learn and pandas, and a deep-learning framework such as PyTorch or TensorFlow.
Familiarity with SQL and working with large datasets.
Curiosity about fraud detection, anomaly detection, or risk analytics, and an approach to understanding a problem domain and translating it into data-driven solutions.
Interest in or exposure to explainable AI and responsible or ethical AI practices.
Strong problem-solving and communication skills, including the ability to explain technical results to non-technical audiences.

Nice To Haves

Exposure to distributed or cloud data tools, such as Spark or cloud data platforms.
Internship, academic project, or coursework experience in machine learning, data engineering, or financial services.

Responsibilities

Explore, profile, and understand transaction, customer, and channel data across wire, ACH, and real-time payment rails.
Develop, train, and validate supervised and unsupervised fraud-detection models, including classification, anomaly detection, and graph or network analysis, for batch and near-real-time use.
Understand major payments fraud categories—including account takeover, authorized push payment scams, synthetic identity, business email compromise, and money mule or laundering patterns—and map them to detection signals and modeling strategies.
Design and maintain behavioral, velocity, device, network, and aggregate feature pipelines, partnering with data engineering to move prototypes into production.
Identify opportunities to apply modern AI techniques, such as deep learning, embeddings, LLMs for unstructured signals, and foundation or graph models, to improve fraud coverage and reduce false positives.
Apply model-interpretability techniques such as SHAP, LIME, counterfactuals, and reason codes to explain flagged payments to fraud analysts, model risk teams, and regulators.
Track model performance, including precision, recall, false-positive rate, and fraud dollars prevented, and present findings and recommendations to technical and business stakeholders.
Partner with fraud subject-matter experts, product, and engineering to take detection ideas from hypothesis through deployment and monitoring.
Follow fraud trends, emerging attack patterns, advances in ML and AI, and responsible-AI practices relevant to banking.
Build skills in programming, mathematics and statistics, distributed computing, and communicating complex results across model science, feature science, and insight science.