AI Dev 25 X NYC | Aditya Dave, John Pepino: Productionizing AI Capabilities in Finance

Why It Matters

BlackRock’s production‑ready AI framework demonstrates how financial firms can unlock client‑centric innovation and internal efficiency at scale while meeting stringent regulatory standards, setting a benchmark for responsible AI deployment in the industry.

Summary

Aditya Dabe and John Pepino of BlackRock opened the session by framing AI as a present‑day necessity for the financial services industry, emphasizing that production‑grade AI solutions are moving beyond experimental prototypes to become core components of client experience and operational efficiency. They highlighted the rapid decline in training and deployment costs, the emergence of mature tooling, and the imperative to embed robust governance and regulatory compliance—particularly under frameworks such as SR 11.7—into every AI deployment to mitigate model risk, citing the 2012 JPMorgan loss as a cautionary tale.

The presenters detailed BlackRock’s three‑pillar AI framework—model integrity and performance, governance and compliance, risk management, and security/privacy—and illustrated how it underpins two primary value streams: client‑facing conversational agents and internal workflow automation. John Pepino walked through the end‑to‑end architecture of a production AI agent, describing a layered moderation pipeline, context‑aware orchestrator, and domain‑driven bots that dynamically select the optimal response path (API, static answer, or LLM) while enforcing guardrails against hallucination, PII leakage, and regulatory breaches. The system’s observability stack tracks latency, embedding generation, vector search, and moderation metrics to pinpoint issues in real time.

The discussion then shifted to Retrieval‑Augmented Generation (RAG) optimization, where Aditya outlined the challenges of hyper‑parameter selection, prompt engineering, and scarce ground‑truth data. He proposed treating RAG pipelines as first‑class ML models—complete with automated ground‑truth generation, memory‑driven retrieval, recursive search loops, and intelligent hyper‑parameter tuning—to accelerate deployment and ensure factual grounding. By systematizing evaluation and embedding RAG within agentic workflows, BlackRock aims to transform static retrieval tools into dynamic, scalable engines that drive both compliance‑safe client interactions and faster quantitative research.

Overall, the session underscored that AI adoption in finance is no longer optional; it is a competitive differentiator that must be balanced with rigorous risk controls. BlackRock’s blueprint—combining regulatory‑aligned governance, modular architecture, and systematic RAG optimization—offers a replicable path for firms seeking to scale AI responsibly while delivering higher‑touch client experiences and operational cost savings.

Original Description

Aditya Dave, Vice President of BlackRock, and John Pepino, Senior Software Engineer at BlackRock, discussed how AI solutions can streamline internal workflows at BlackRock, such as automated compliance checks or portfolio rebalancing.

At AI Dev 25 x NYC, they discussed why the financial services industry is at an inflection point where AI is a present-day necessity, used by banks, insurers, and asset managers to improve efficiency and client experience.

----------

Join us at AI Dev 26 x San Francisco! Tickets: https://ai-dev.deeplearning.ai