🤖 Autonomous Financial Research Agent

A fully autonomous multi-agent system that conducts end-to-end financial market research — from raw company name to investor-ready report — with zero human intervention.

---

⚡ Results at a Glance

Metric	Detail
🤖 Agents	2 specialized (Senior Financial Researcher + Technical Financial Writer)
🔄 Orchestration	Process.sequential — Writer is blocked until Researcher completes
🎯 LLM	gemini/gemini-2.5-flash-lite via CrewAI's LLM wrapper
⚙️ Dynamic Input	Single {company} variable — swap any ticker, no code changes
📝 Output	3-paragraph investor-ready Markdown report with BUY/HOLD/SELL thesis
🧪 Test Case	NVIDIA — growth drivers, risk factors, and recommendation generated autonomously

---

🧠 Why Multi-Agent Over a Single Prompt?

A single LLM prompt produces shallow, generic financial summaries. This system separates research from writing into two distinct agent personas — each with its own role, goal, and backstory. The quality difference is significant:

Single Prompt:       "Analyze NVIDIA" → Generic 2-paragraph summary

This System:         Researcher Agent  → Identifies 3 growth drivers + 3 risks
                           ↓ (sequential handoff — writer waits for validated output)
                     Writer Agent      → Transforms research into structured
                                         investor-grade Markdown report

Role specialization plus sequential enforcement means the Writer never hallucinates research it hasn't received.

---

🏗️ Agent Architecture

# Agent 1: Researcher
researcher = Agent(
    role='Senior Financial Researcher',
    goal='Uncover deep insights into {company} and its 2025 market outlook.',
    backstory="World-class financial analyst specializing in emerging market trends.",
    llm=gemini_flash
)

# Agent 2: Writer
writer = Agent(
    role='Technical Financial Writer',
    goal='Synthesize research findings into a clear, investor-ready report.',
    backstory="Seasoned business journalist transforming raw data into executive narratives.",
    llm=gemini_flash
)

# Sequential crew — researcher must finish before writer starts
fin_crew = Crew(
    agents=[researcher, writer],
    tasks=[research_task, writing_task],
    process=Process.sequential
)

---

🛠️ Tech Stack

Layer	Technology
Multi-Agent Framework	CrewAI (Agent, Task, Crew, Process)
LLM	gemini/gemini-2.5-flash-lite (via CrewAI LLM wrapper)
Orchestration Pattern	Process.sequential — strict task dependency enforcement
Environment	python-dotenv for secure API key management
Language	Python 3.10+

---

📊 Live Execution: NVIDIA Case Study

Input: fin_crew.kickoff(inputs={'company': 'NVIDIA'})

The Researcher agent autonomously identified:

Growth Drivers identified:

• AI/Accelerated Computing dominance (Blackwell GPU architecture)
• Automotive sector expansion (NVIDIA DRIVE platform)
• Sustained gaming ecosystem revenue (RTX/DLSS)

Risk Factors identified:

• Hyperscaler custom silicon competition (Google TPU, AWS Trainium)
• TSMC supply chain concentration risk
• Semiconductor industry cyclicality

Writer output: A polished 3-paragraph Markdown investment summary with a BUY/HOLD/SELL recommendation — formatted for executive stakeholders.

BUY Scenario	HOLD Scenario	SELL Scenario

---

🔑 Key Engineering Decisions

• Why Process.sequential? The Writer agent must receive validated research before drafting. Sequential enforcement prevents the Writer from generating fabricated analysis — a critical guardrail in financial contexts.
• Why role + backstory per agent? CrewAI agents perform significantly better with a defined persona. The "world-class analyst" backstory steers the Researcher toward structured, data-driven output; the "journalist" backstory pushes the Writer toward narrative clarity.
• Why {company} variable injection? One-line change to kickoff(inputs={'company': '...'}) re-targets the entire pipeline to any publicly traded company — no prompt editing, no code changes.

---

🚀 Quick Start

# 1. Clone
git clone https://github.com/Rahilshah01/autonomous-financial-research-agent.git
cd autonomous-financial-research-agent

# 2. Install
pip install crewai python-dotenv

# 3. Set API key
echo "GEMINI_API_KEY=your_key_here" > .env

# 4. Run (change company in kickoff inputs if desired)
python main.py

---

📁 Repository Structure

autonomous-financial-research-agent/
├── main.py          # Agents, tasks, crew definition + kickoff
├── images/          # Sample output screenshots (buy/hold/sell)
├── .env.example
├── requirements.txt
└── README.md

---

Built by Rahil Shah · MS Data Science @ Stevens Institute of Technology