Website visitor prediction

We were approached by a German software development company to develop a customer journey prediction AI for a Southern German automotive company. The end customer was interested in how well a modern AI system could predict the next website a user would visit based on their browsing history. We used a dataset of millions of past website visits to train a modern language model.

Project Details

Industry: Automotive
Time Frame: 2023
Technologies / Products used:

• Apple MPS acceleration for Neural Network Training
• fast.ai, PyTorch and Pandas frameworks
• Google Cloud Firebase for cloud model deployment
• REST API for UI - Backend Interface
• Jupyter Notebooks for fast and agile model development and training

{"prediction": 
    ["car-configurator.select_motorization", "models", "product_list.cars"], 
    "time": 92}