Leads engineering for the AI Servicing transformation of Aladdin, BlackRock’s investment platform.
Session
Main Stage
Date
Monday, November 30
Level
Everyone
// AI Engineering Skill /
Building agentic systems
// 01 / Bio
Sankara Ramakrishnan is a Senior Principal Engineer at BlackRock, where he leads engineering for the AI Servicing transformation of Aladdin, the firm’s investment platform. His team owns the Intelligent Servicing Platform, which takes in client requests and hands each one to a domain-specific agent that can answer the question or carry out the task on Aladdin.
Before moving into this role, Sankara was an architect for RockAI, BlackRock’s internal AI platform, which supplies the agent development kit and model inference his team now builds on. His work today covers agent architecture, knowledge-graph systems, and guardrails designed for what each agent does, so that agents can operate safely in a regulated industry.
He joined BlackRock in 2021 after more than seven years as a principal architect at MetLife. There he designed a cloud data platform for investment management, built out the company’s global DevOps practice, and was technical lead on moving a trading operation with more than $400 billion in assets onto Aladdin.
// 02 / The session
Guardrails for client-facing agents in a regulated firm
Abstract
What does it take to put AI agents in front of clients at one of the world’s largest asset managers? BlackRock engineers share how they build guardrails into client-facing agents on Aladdin, in an industry where every answer has to hold up to scrutiny.
Client requests on Aladdin go through the Intelligent Servicing Platform, which hands each one to a domain-specific agent that can answer the question or carry out the task. A generic content filter doesn’t know what a given agent is supposed to do, so the team designs guardrails around each agent’s specific job.
Sankara and Eswaran walk through how these agents are built on RockAI, BlackRock’s internal agent development kit and inference platform, how they’re monitored once they’re live, and what it took to get generative AI approved and governed inside a regulated financial firm.
Co-presented with Eswaran Rathinam.
What you’ll learn
Why generic guardrails fall short for domain agents
How BlackRock builds client agents on its internal platform
Getting generative AI approved in a regulated firm
AI DEV · BY DEEPLEARNING.AI
Join us at AI Dev
Meet the engineers shipping AI to production, and leave ready to do it yourself.