Solance AI
Overview
Solance AI is a web platform for cryptocurrency visualization and short-term trend prediction (5-minute candle forecasting). The system integrates automatic training and deployment of AI models designed to assist in high-frequency decision-making.
What's the purpose of this project ?
Financial decisions in crypto trading are often driven by intuition rather than data, leading to inconsistent performance and higher risk. Solance AI aims to make short-term investment decisions data-driven by providing real-time predictions, visual insights, and model-based recommendations — empowering users to invest with confidence.
How can it be done ?
Implemented with Python (FastAPI) for the backend, React + Tailwind for the frontend, and AWS + Docker for deployment. Models trained using Scikit-learn, PyTorch, and TensorFlow for prediction tasks; live data processed through WebSocket APIs for real-time updates.
How did I do it ?
The solution combines a forecasting engine powered by trained ML models and a dynamic dashboard for real-time visualization. The system fetches live crypto data, preprocesses it, runs predictions through deployed models, and displays key indicators through interactive charts — with automated deployment through CI/CD and MLOps practices.
Results
The models achieved positive and robust results, averaging 7% monthly ROI on real historical data, demonstrating strong predictive performance across multiple coin pairs. Automated retraining and near real-time inference were validated at trading-scale windows, confirming the feasibility of an end-to-end deployment pipeline. Ongoing experiments aim to dynamize risk management modules using OptNets and unfolded layers, enhancing adaptability to changing market conditions and improving portfolio-level risk control. Development is currently on pause due to academic priorities, with only a few technical components remaining: frontend/UI refinements, containerization and automation for model retraining, and production deployment polish. This project highlights a complete workflow from quantitative modeling and optimization to backend orchestration and frontend visualization, demonstrating strong capabilities in ML, operations research, and full-stack deployment.