EdgeTrader AI

Main Page

EdgeTrader AI

🚀 Inspiration

I wanted to build an intelligent trading assistant that doesn’t just output BUY or SELL signals, but actually reasons like a human analyst. Most trading tools only show indicators — I wanted a system that explains why. This led to building an agentic AI that blends ML signals, price action, sentiment analysis, and an LLM reasoning layer.

📈 What it does

EdgeTrader AI generates swing-trading recommendations using:

RandomForest model predictions
Candlestick pattern recognition
Sentiment scoring
Dynamic holding horizon selection
Monte-Carlo forecasting

It also uses an LLM layer to evaluate market context and deliver human-like reasoning for every recommendation.

Currently, it is tuned for NVDA, but the architecture is designed to expand to all stocks and eventually manage entire trading portfolios.

🛠️ How we built it

Backend: FastAPI powering ML inference, pattern extraction, sentiment analysis, trend scoring, and simulation forecasts
Frontend: React + ApexCharts for candlestick charts, price projections, and agentic insights
LLM Layer: Local Ollama models for contextual evaluation and natural-language reasoning
ML Models: RandomForest-based BUY/HOLD/SELL classifier trained on NVDA historical data

⚠️ Challenges we ran into

Keeping LLM outputs consistent and non-hallucinatory
Managing complex React state across multiple async endpoints
Extracting meaningful patterns from noisy candlestick data
Designing a clean UI despite many insights
Ensuring responsiveness while running multiple models

🏆 Accomplishments that we're proud of

Blending ML forecasts, sentiment, price patterns, and LLM reasoning in one system
Achieving interpretable explanations rather than black-box outputs
Building a smooth, modern interface with data + reasoning visualized
Designing a modular pipeline that scales beyond NVDA