The New York edition of Data Science Salon offered a mix of tech and business talks on AI and finance.Experts discussed the latest advances that can be applied in financial services, with caution and emphasis on the highly regulated environment the sector faces daily.
Between sessions, participants had access to a networking space to debate the topics.
Unlocking the Potential of Large Action Models
Kristi Baishya of Nomura opened by introducing Large Action Models. “Large Action Models can not only understand language, but also take actions,” she said.
LAMs plan and execute multi-step decisions; she highlighted deployments in finance, healthcare, and supply-chain ops, and covered reliability, ethics, and compute budgets.
Data Science in Stress Testing
Akhil Khunger, VP of quantitative analytics at Barclays, shifted the focus to risk. “It’s crucial to understand what stress testing actually looks like,” he said. “Financial institutions must conduct it to evaluate whether they have adequate capital resilience.”
By blending historical market shocks with forward-looking scenarios, he showed how ML-driven tests reveal hidden balance-sheet weaknesses before regulators - or crises - arrive.
Simplifying AI Interactions for Users
Kieran Gallagher of S&P Global EDM argued the last mile of AI is human. Adding “humans-in-the-loop” workflows and richer semantic catalogs turns desktop AI tools into intuitive copilots that support natural-language queries and data reconciliation.
“All these data-quality processes don’t tell us whether we got what we need. That’s where data management comes in,” he said. “We must evaluate both content and accuracy.”
From Chaos to Clarity: Streamlining Data for Better Decision-Making
Yashasvi Singh, data steward at Navy Federal Credit Union, outlined a plan for taming messy data estates: standardized governance, automated QA scripts, and a living data catalog that lets analysts self-serve without version-control issues.
“There’s excitement about generative and agentic AI, but if we feed these systems flawed data, we get flawed answers - garbage in, garbage out,” she said.
SqPal: GenAI Text-to-SQL Tool for Product Analytics
Dan Liyanage of PayPal detailed building a GenAI tool that converts natural language to production-ready SQL, covering secure rollout, feedback loops, and gaps in metadata.
The Landscape of Risk Models in Equities
Arkin Gupta, VP at Morgan Stanley, surveyed equities risk modeling, from classic factor models to GPU-heavy deep nets, weighing interpretability against predictive power and fitting large models into real-time trading.
Panel: Building & Scaling Gen AI in Financial Services
Leads from S&P Global, JPMorgan Chase, T. Rowe Price, Capital One, and American Express shared lessons on moving Gen AI prototypes to production, covering latency, governance, and whether RAG pipelines or custom copilots win budgets.
Gen AI Use Cases in Asset Management & Lessons Building a Copilot
Yu Yu, director of Data Science at BlackRock, mapped how Gen AI boosts client engagement and research, then dissected the RAG-plus-Agents stack behind BlackRock’s client copilot and its post-launch monitoring.
“Gen AI can draft materials, but humans make final decisions on what clients see,” she said.
Using GenAI to Boost Developer Engagement, Learning, and Efficiency
Harry Mendell, Data Architect at the Federal Reserve Bank of New York, showed Gen AI as team translator - explaining legacy code, suggesting refactors, and bridging gaps between COBOL custodians and cloud-native developers.
“Agents aren’t just translating code; they’re also educators who explain what our tools can do,” he said.
Summary
The event offered a closer-knit setting to meet AI professionals in finance without the flash of larger conferences.
Because of strong interest, Data Science Salon New York will run twice this year, with a second edition on December 11. Sessions from May are available on demand.
