An AI-powered system that analyzes startup ideas using real-world startup data, not just LLM intuition.
VentureLens is a data-driven startup analysis tool that evaluates business ideas by:
- retrieving similar startups from a real dataset
- computing evidence-based signals
- scoring viability and risk
- generating structured future scenarios
- producing grounded recommendations
Unlike typical AI tools, VentureLens does not rely on LLM guessing — it is built around a Retrieval-Augmented Generation (RAG) pipeline with real startup data.
User Idea ↓ Similarity Retrieval (TF-IDF) ↓ RAG Context Builder ↓ Scoring Engine (data-driven) ↓ Risk Analyzer ↓ Scenario Simulator ↓ Recommendation Engine ↓ Report Generator ↓ Streamlit UI
- Finds similar startups from dataset
- Uses similarity scores to ground reasoning
- Peer success score
- Similarity confidence
- Evidence quality
- Risk penalty calibration
- Detects weak market patterns
- Identifies failure signals from peers
Each scenario includes:
- description
- why it could happen
- key trigger
- warning signs
- strategic actions
Multi-dimensional startup evaluation:
- Viability
- Retrieval Strength
- Peer Quality
- Evidence Depth
- Risk Control
The system uses a processed startup dataset containing:
- description
- industry / sub-industry
- funding (total_funding_usd)
- success_score
- outcome_label
This enables evidence-based reasoning instead of hallucination.
git clone https://github.com/yourusername/venturelens-ai
cd venturelens-ai
pip install -r requirements.txt
streamlit run app.py
## 🧪 Example Use Cases
- Validate startup ideas before building
- Practice product thinking for interviews
- Analyze why similar startups succeeded or failed
- Explore market patterns in different industries
## 🧠 Design Philosophy
- “Don’t let the model guess — force it to reason from data.”
- VentureLens separates:
Data layer → similarity + signals
Logic layer → scoring + risk rules
Narrative layer → explanation
This makes the system:
more interpretable
more realistic
more useful for learning
## 📌 Future Improvements
- Replace TF-IDF with embedding-based retrieval
- Add industry baseline scoring
- Improve calibration using percentile ranking
- Integrate real-time startup APIs
- Add memory & history tracking
## 💡 Why This Project Matters
- Most AI startup evaluators are just chatbots.
- VentureLens is different:
- grounded in real data
- modular architecture
- explainable outputs
- This makes it closer to a real decision-support system than a demo app.
## 👨💻 Author
- Built as a hands-on AI systems project
- Focus: turning AI from “text generator” → “reasoning system”