What Changed
From Scattered Signals to One Graded Recommendation

The goal was a recommendation an investor could act on, with the reasoning visible behind it. Here is how Athena differs from the manual process it replaces.

User Research

Single Image
How We Work

From Analyst Workflow to a Coordinated Agent Pipeline

1
Problem framing & pipeline design

Mapped how a real analyst works onto distinct agent roles: retrieval, research, analysis and reporting.

2
Data & tooling integration

Connected live market data, company fundamentals, filings and news sources so every analysis rests on current information.

3
RAG knowledge layer

Built a Qdrant-backed retrieval layer over source documents like SEC filings and earnings reports.

4
Multi-agent orchestration

Built the LangGraph-based deterministic workflow linking the retrieval, research, analysis and reporting agents.

5
Analysis engine

Implemented parallel technical (RSI, MACD, Bollinger Bands) and fundamental (DCF) models feeding directly into reporting.

6
Frontend, streaming & tracking

Built the React interface that streams each agent’s reasoning live and keeps watching tracked portfolios afterward.

Software & Tools

The Tools Behind the Innovation