deepblueai/sigir_ecom_2021_data_challenge_1st_place ? reverse-engineered prompt

Build me a Python project that recreates the winning solution for the SIGIR eCOM 2021 Data Challenge.

I want two parts in it, one for session based product recommendation, where the model looks at the first few events in a session and predicts the later product interactions, and another for intent prediction, where it sees a session with an add to cart event and predicts whether that item gets bought before the session ends.

Please make it easy to run, with clear scripts or notebooks for training and prediction, and simple instructions for preparing the competition data and reproducing results. If it helps, organize the recommendation part and the intent prediction part separately, but keep the overall project straightforward to follow. Add anything useful for evaluation and generating submission files too.

If you need to check the original competition details or data format, look them up online and match the expected setup as closely as possible.