Shenson Joseph, Senior AI, ML, and Data Engineer at JPMorgan Chase & Co., speaks with Jessie Smith, VP of Product Management at Ataccama, in a video interview about the role of data quality and unstructured data in AI, opportunities and challenges with GenAI in financial services, and the implementation of AI in different sections.

Beyond his current corporate role, Joseph is an AI researcher and data scientist with expertise in quantum computing, specializing in machine learning innovations, explainable AI, and AI-driven anomaly detection systems.

He holds dual master’s degrees in data science and electrical engineering and is now pursuing a Ph.D. at the University of North Dakota. His research centers on the integration of quantum computing and AI.

Joseph’s previous work at Nokia involved spearheading innovations in financial risk modeling, fraud detection, and cybersecurity using AI. He also serves as a board member for Photon IV, a Canadian startup that guides innovations in optical satellite ground stations.

Managing data quality in financial AI systems

When asked to share strategies for improving data quality and accuracy, Joseph says, “Data quality is a backbone of AI, especially in the financial industry, where smaller rates can have big consequences.”

Given the high stakes involved, he advocates for a rigorous approach to managing data quality and outlines a multi-pronged strategy that combines:

• Automated data validation
• AI-powered anomaly detection
• Strong data governance policies

Beyond validation and detection, Joseph emphasizes transparency throughout the data lifecycle. “It’s also important to ensure data lineage and transparency so everyone knows where the data comes from and how it’s being used,” he adds.

Reaping actionable insights from unstructured data

Moving forward, Joseph says, “Unstructured data is the backbone of our AI research.” He adds that unstructured data in the financial industry, such as contracts and customer communication, holds a wealth of insights.

He mentions how Natural Language Processing (NLP) models can analyze text, extract key entities, and detect sentiment while summarizing complex documents.

For example, Joseph highlights the massive implementation of RAG models to summarize certain complex documents. AI can scan legal contracts to flag risky clauses or automate customer service analysis to identify emerging issues before they escalate.

Opportunities and challenges with GenAI

Labeling GenAI as a game changer, Joseph maintains that it can personalize financial advice, automate regulatory reports, and even generate synthetic fraud scenarios for better anomaly analysis.

Highlighting challenges, Joseph says, “Model hallucination, regulatory concerns, and ethical considerations need to be tackled carefully to ensure reliability and trust.” He also mentions the growing trend of agentic AI systems and calls out the ethical challenges in that domain and the need for solid regulations.

Regarding the implementation of AI in the financial services and telecom sectors, Joseph notes that both sectors rely on real-time data. However, telecom focuses mostly on network optimization, and finance is more about compliance and risk.

Drawing from his background in the telecom industry, Joseph explains that much of the work centered around predictive maintenance can also be effectively translated into finance for purposes like fraud detection and proactive risk management. In telecom, AI techniques and machine learning models have been extensively utilized to optimize network performance and enhance user experience.

CDO Magazine appreciates Shenson Joseph for sharing his insights with our global community.