Back to Speakers
Soumith Javvaji - Lead AI Engineer at Transient AI

Connect

November 23–27, 2026

Soumith Javvaji

Lead AI Engineer

Transient AI

Session Details

From Headlines to Holdings: A Multi-Agent System That Explains Market Impact on Portfolios

When markets move, financial advisors need answers fast: which client portfolios are affected by this rate decision, this filing or this headline, and why? Doing that manually means scanning news, regulatory filings and macroeconomic data, then mapping it back to individual holdings. In this session, I walk through a multi-agent system I built for wealth management that does this automatically. The system splits the problem across specialized agents orchestrated with LangGraph: an ingestion layer that continuously pulls market news and data from sources such as SEC filings, Federal Reserve macroeconomic releases, financial news and GDELT; a research agent that uses RAG to ground its reasoning in those sources; and an analysis agent that maps events to the holdings in a portfolio and explains the likely impact, with citations an advisor can check. I'll cover the design decisions that made it work in practice: - How to split responsibilities across agents, and when a single agent is enough - Keeping real-time ingestion fast and affordable with caching - Grounding every explanation in a source, so advisors can verify it instead of trusting it blindly - Evaluating tool calls and tracing each step to catch wrong tools, stale data and ungrounded claims before they reach a client conversation Attendees will leave with a reusable architecture for event-driven, multi-agent analysis that applies well beyond finance.

Full conference agenda

See every session across the AI Agents Conference November 2026 line-up.

View Agenda