VC Investment Decision Support System

A multi-agent decision-support service for venture capital investment analysis. This system evaluates startups across multiple dimensions—market opportunity, financial health, founder quality, and probabilistic business survival—to provide a comprehensive investment recommendation.

Note: This is a decision-support tool. It informs, but never replaces, human investment judgment.

🏗️ System Architecture

The backend is composed of four distinct AI/analytical agents, each exposed as a FastAPI microservice. A demo pipeline script aggregates their outputs to formulate a final investment decision.

• Agent 1: Market Intelligence (Port 8002)
• Agent 2: Financial Health (Port 8003)
• Agent 3: Founder Scoring (Port 8000)
• Agent 4: Monte Carlo Simulation (Port 8001)

🚀 Quick Setup & Usage

Prerequisites

• Python 3.10 or newer

Installation

1. Open a terminal in the project folder.
2. Install the required dependencies:

   pip install -r requirements.txt

Running the System

1. Start the Agent Servers: In the terminal, run the following command and leave the window open until you see "ALL AGENTS RUNNING!":

   python run_all.py

   (This script starts all four FastAPI agents in the background.)

2. Run the Analysis Pipeline: Open a second terminal in the same folder and execute the demo pipeline to see the full analysis:

   python demo_pipeline.py

3. To Stop: Go back to the first terminal and press CTRL+C.

---

🔍 Detailed Agent Breakdown

Agent 1: Market Intelligence Scoring (agent1_market.py)

Endpoint: POST /market-score (Port 8002)

Scores a startup's market opportunity based on four key dimensions:

• Market Size
• Traction
• Competition
• Timing

Key Features:

• Uses a curated mock knowledge base of realistic industry profiles (TAM, growth, maturity, competitors). In production, this would integrate with Crunchbase, PitchBook, Gartner, etc.
• Returns an evidence-backed market report including green flags, red flags, market growth rate, and competition intensity (which feeds directly into Agent 4).

Agent 2: Financial Health Scoring (agent2_financial.py)

Endpoints: POST /financial-score (JSON) | POST /financial-score/csv (CSV upload) (Port 8003)

Evaluates the startup's financial viability and unit economics.

• Metrics Analyzed: Monthly Burn Rate, Runway (in months), Unit Economics (LTV/CAC ratio), and Revenue Growth.
• Key Features:
  • Extracts actionable red/green flags based on thresholds (e.g., LTV > 3x CAC, declining revenue, less than 12 months runway).
  • Prepares a clean data payload (agent4_payload) to seamlessly feed into the Monte Carlo simulation.

Agent 3: Founder Success Scoring Engine (agent3_founder_score.py)

Endpoint: POST /founder-score (Port 8000)

Evaluates founder quality and returns a comprehensive Founder Success Score (0-100).

• Evaluation Criteria:
  1. Psychometric Assessment (SFSS): Evaluates 6 dimensions (Resilience, Opportunism, Curiosity, Courage, Innovation, Leadership) using rule-based evidence scoring.
  2. Track Record (Founder Level): Classifies founders from L1 (First-Time) to L10 (Proven Builder) based on exits, experience, and scale.
  3. Persona Classification: Uses nearest-prototype cosine similarity matching (e.g., "Serial Entrepreneur", "Tech Executive", "Product Visionary").
  4. Statistical Prediction: Utilizes a Random Forest / XGBoost model trained on a synthetic VCBench dataset to predict success probabilities based on education, experience, funding history, GitHub activity, patents, etc.

Agent 4: Business Survival Monte Carlo Engine (agent4_monte_carlo.py)

Endpoint: POST /simulate-business (Port 8001)

Simulates the startup's financial future under uncertainty by combining outputs from Agents 1, 2, and 3.

• Key Features:
  • Runs 10,000 month-by-month scenarios simultaneously using heavily optimized vectorization (NumPy).
  • Outputs Include:
    • Survival probabilities at 12, 24, and 36 months.
    • Failure-mode breakdown (e.g., "Burn rate too high", "Customer churn too high").
    • Cash-crunch timing distributions.
    • Key drivers of success/failure via Spearman correlations.
    • Stress tests (best case, worst case, realistic case).
    • Founder impact metrics (score differences between succeeding and failing scenarios).
    • Actionable recommendations, including counterfactual scenarios (e.g., "Reducing burn by 15% improves Year-1 survival...").

---

🧪 Demo Pipeline (demo_pipeline.py)

The pipeline script acts as the orchestrator/client for the system. It simulates a VC analyst requesting an end-to-end report on a mock startup ("Clarify Health").

Workflow:

1. Verifies that all 4 agent servers are healthy and reachable.
2. Calls Agent 1 for the Market Score.
3. Calls Agent 2 for the Financial Score.
4. Calls Agent 3 for the Founder Score.
5. Aggregates data from Agents 1-3 and submits it to Agent 4 for Monte Carlo simulation.
6. Computes a Weighted Composite Score (30% Market, 40% Financial, 30% Founder).
7. Outputs a beautiful CLI report including scores, red/green flags, probability metrics, and a Final Investment Decision (e.g., STRONG INVEST, CONSIDER, INVESTIGATE FURTHER, PASS).