Conversations with Frank Fabozzi, CFA, Featuring Iro Tasitsiomi, PhD

Why It Matters

Integrating AI responsibly into investment processes can unlock superior insights while safeguarding against model risk, making finance teams more agile and data‑driven.

Key Takeaways

• Physics background drives abstraction and data efficiency in finance models.
• AI value hinges on alignment with investment philosophy and culture.
• NLP unlocks unstructured data for thematic investing and sentiment analysis.
• Model choice depends on problem definition, data quality, and interpretability.
• Credibility requires finance knowledge alongside technical AI expertise.

Summary

The interview with Iro Tasitsiomi, head of AI and investments data science at T. Rowe Price, explores how a physicist’s training shapes AI strategy in asset management. Her transition from cosmology research to quantitative finance illustrates the power of abstraction, data efficiency, and scientific skepticism when building investment models.

Tasitsiomi emphasizes that AI initiatives must first maximize value and align with the firm’s investment philosophy and culture. She highlights natural‑language processing as a game‑changer for extracting insights from unstructured text, enabling thematic investing, sentiment tracking, and rapid identification of niche exposures. Model selection, she explains, hinges on problem definition, data richness, and the trade‑off between performance and interpretability, advocating for parsimonious solutions that avoid unnecessary complexity.

Concrete examples include using NLP to monitor global news for thematic signals and a rapid ChatGPT‑driven simulation of an S&P 500 index reconstitution. She warns against opaque “black‑box” models, recalling the Gaussian copula’s failure, and stresses the need to “poke” models to understand behavior under stress. Her scientific rigor translates into a disciplined, skeptical approach to AI deployment.

For asset managers, the discussion underscores that successful AI adoption requires not only technical skill but deep finance fluency to earn credibility with portfolio managers. Aligning AI tools with business objectives, fostering cultural acceptance, and maintaining model transparency will determine whether AI delivers sustainable investment edge.

Original Description

As artificial intelligence becomes increasingly embedded in investment workflows, distinguishing meaningful insights from unnecessary complexity has become an important challenge for investment leaders.

Join us for a timely discussion with Iro Tasitsiomi, PhD, an astrophysicist and investment data science leader, as she explores how to use AI and machine learning thoughtfully while preserving human judgment, transparency, and risk control.

In conversation with Frank Fabozzi, CFA, Tasitsiomi explains that sophisticated models add value only when paired with well-defined problems and high-quality data. Echoing Einstein, she argues that models should be made as simple as possible, but not simpler. She also shares when advanced machine learning techniques are most effective, particularly for capturing nonlinear relationships, working with unstructured data, or enabling unsupervised discovery.

Key Discussion Points:

Principle of parsimony: When the simplest model preserves essential insights and helps avoid unnecessary complexity, cost, and risk.

Poke the model: Why probing model behavior, especially outside normal conditions, helps investment teams understand potential risks and failure modes.

Understand the tools: Why sound judgment about data, assumptions, and model limitations matters more than coding skills, especially for younger analysts building credibility on AI-enabled investment teams.

Speak the language: Why successful AI adoption requires both technical rigor and clear communication across investment teams.