I need a data-driven workflow that pulls air-quality readings from reliable public APIs, cleans and aggregates them, and then forecasts future levels of NO2 and CO. My priority is an LSTM-based model, though I’m open to brief comparisons with simpler baselines if it helps demonstrate the benefits. What I expect: • API integration script that automatically downloads the historical and latest measurements. • Clear preprocessing and feature-engineering steps to handle missing values, seasonality, and weather or temporal variables you feel add value. • An LSTM model implemented in Python (TensorFlow or PyTorch preferred) with well-commented code, training routine, and hyper-parameter notes. • Evaluation plots and metrics on a held-out test set, plus a short explanation highlighting how far ahead the model can reliably predict. • Reproducible Jupyter notebook (or equivalent) and a concise technical report summarizing findings, assumptions, and next-step recommendations. If you’ve previously combined ARIMA, Prophet, or similar techniques with deep learning, feel free to suggest how that might refine the forecasts—but keep the core delivery focused on LSTM. Clean code, clear documentation, and results I can replicate on my machine will be the key acceptance criteria.