Inspiration
Financial markets generate massive amounts of data every second, making it difficult for traders to separate meaningful signals from market noise. Most retail traders either rely on basic indicators or expensive institutional tools.
We wanted to build an intelligent AI assistant that combines advanced forecasting models with explainable AI to make professional-grade market intelligence accessible to everyone.
What it does
ATM analyzes financial markets by combining:
Historical price data Technical indicators Market sentiment News analysis Macroeconomic events AI reasoning
The platform then provides:
AI-generated market analysis Probabilistic price forecasts Trend detection Risk assessment Support & resistance identification Explainable predictions Natural language trading insights
Instead of saying:
"Buy EUR/USD."
ATM explains:
"EUR/USD has a 74% probability of continuing its bullish trend over the next 24 hours due to increasing buying pressure, positive macroeconomic sentiment, and momentum confirmation across multiple indicators
How we built it
The platform is designed around a modern AI architecture.
Frontend
Next.js React Tailwind CSS TypeScript
Backend
FastAPI Python
AI & Machine Learning
OpenAI GPT API PyTorch XGBoost Time-series forecasting models Ensemble learning
Data Processing
Financial market APIs Technical indicator engine Feature engineering pipeline
Infrastructure
PostgreSQL Redis Docker Cloud deployment
Challenges we ran into
One of the biggest challenges was balancing prediction performance with explainability.
Financial markets are highly dynamic and influenced by countless variables, making perfect prediction impossible. Instead of attempting to guarantee outcomes, we focused on generating probability-based forecasts supported by transparent reasoning.
Another challenge was designing an architecture that combines machine learning forecasting models with large language models while keeping the system scalable, efficient, and easy to extend.
Accomplishments that we're proud of
Designed an explainable AI architecture instead of a black-box prediction engine. Combined AI reasoning with quantitative market analysis. Built a scalable foundation capable of supporting multiple financial markets. Focused on helping users understand market behavior rather than simply producing signals.
What we learned
This project reinforced an important lesson: successful financial AI is not only about prediction accuracy.
Building useful market intelligence requires combining statistical forecasting, domain knowledge, explainability, and user trust. We also learned the importance of designing AI systems that clearly communicate confidence, uncertainty, and the reasoning behind every insight.
What's next for ATM
Our roadmap includes:
Support for Forex, Stocks, Crypto, Commodities, and Indices Real-time market data integration AI-powered news and sentiment analysis Personalized trading insights Portfolio analysis and risk management Reinforcement learning research for adaptive forecasting Multi-agent AI collaboration Mobile applications Enterprise APIs for financial institutions
Ultimately, our vision is for ATM to become an intelligent financial co-pilot that empowers traders and investors with transparent, AI-driven market intelligence.
Built With
- docker
- fastapi
- github
- nextjs
- numpy
- openai
- pandas
- polygon
- postgresql
- python
- pytorch
- redis
- render
- scikit-learn
- typescript
- vercel