Stock Market Predictive Analytics AI

I need a compact yet accurate predictive-analytics engine that focuses exclusively on stock-market behaviour. The goal is to ingest historical price data, basic fundamentals and real-time market feeds, then return probability-based forecasts for short- and medium-term price movement.

You are free to choose the modelling approach—classical time-series, machine-learning ensembles, or deep-learning (e.g., LSTM, Transformer)—as long as the final model outperforms a naïve benchmark and can be retrained with fresh data. Python is preferred because I already license data through a Python API, but I’m open to R or Julia if you can wrap it in a simple REST endpoint.

To keep scope clear, here is what I expect as concrete deliverables:
• Data-preparation script that cleans and aligns OHLCV, splits train/validation/test, and logs data quality issues.
• Reproducible model notebook or .py file with well-commented code, hyper-parameter settings and evaluation metrics (MAE, RMSE, Accuracy on directional move).
• A lightweight API or CLI that takes a ticker symbol and date range as input and returns the forecast plus confidence score.
• Short README explaining installation, retraining and expected hardware requirements.

I’ll provide sample tickers and the data-vendor credentials once we agree on an approach. Your proposal should briefly outline the modelling technique you favour and a timeline for first results; code style and clarity will be part of the acceptance criteria.