Building an Open-Source AI Agent That Makes Financial News Understandable

Financial news is easy to access, but understanding it is not.

FinLens is an open-source AI agent that explores a different approach to financial news: instead of generating another summary, it tries to determine what a person actually needs to understand about an event.

The project automatically gathers financial news from external sources and gives an open-weight LLM access to tools for retrieving and checking information. The agent then turns selected events into short, beginner-friendly explanations. Depending on the story, the explanation can be a cause-and-effect flow, a simple analogy, a financial term explained in context, or a connection to an everyday situation.

For example, instead of producing a paragraph about rising oil prices, FinLens can explain the chain:

Oil prices rise
→ transport costs increase
→ businesses face higher costs
→ some everyday prices may rise

The talk will cover:

• The problem with treating AI financial applications as simple news summarizers

• The architecture of FinLens: news sources, retrieval tools, agent, LLM and presentation layer

• How tool calling allows the agent to gather additional context instead of relying only on a model's existing knowledge

• How the agent decides what kind of explanation a particular story needs

• Designing AI output that is short, visual and useful instead of producing long blocks of generated text

• Using an open-weight model through Ollama and the trade-offs between local and cloud inference

• Practical challenges encountered while building and testing the application on limited local hardware

• What worked, what did not, and what I would change when building an agent-based application from scratch

The session will include a live walkthrough of FinLens and its architecture, with examples of the same type of financial event being transformed from a news headline into an explanation a beginner can actually follow.

The goal is not to present a finished "AI news app", but to share the engineering process and lessons from building an open-source AI agent around a real-world understanding problem.

• Understand the architectural difference between an LLM-powered application and an AI agent that can use external tools.

• Learn how FinLens combines news retrieval, tool calling and an open-weight LLM to build a practical application.

• See how the same financial event can be transformed into different explanation formats depending on what a beginner needs to understand.

• Learn practical techniques for designing AI output that is concise, structured and useful rather than simply generating more text.

• Understand the trade-offs between local and cloud inference when building with open-weight models on limited hardware.

• Take away practical lessons from building, testing and open-sourcing an AI agent from scratch.