The Applied Scientist will develop and improve time series forecasting models for Amazon Devices sold through offline channels, with a specific focus on addressing the cold-start problem for new, unlaunched devices. They will implement and refine machine learning models that can accurately forecast demand patterns across different color variations of new Amazon devices, helping optimize inventory and assortment decisions for brick-and-mortar operations before product launch. This role offers an excellent opportunity to tackle complex forecasting challenges where historical data is limited or non-existent, requiring innovative solutions to capture potential demand signals and color-specific preferences for unreleased products..
Key job responsibilities
- Develop and implement forecasting models using Python, focusing on techniques that address the cold-start problem for new Amazon Devices, including transfer learning from similar products and incorporating external factors that may influence color-mix demand.
- Apply deep learning frameworks (TensorFlow/Keras, PyTorch) to build sequence models (RNNs, LSTMs) that can effectively utilize limited pre-launch data and proxy information to forecast color-specific demand for new devices.
- Conduct sophisticated feature engineering tasks, leveraging data from product attributes, market trends, and consumer preferences to create informative features for forecasting demand of unlaunched devices across different color options.
- Evaluate model performance using appropriate metrics for cold-start scenarios, create visualizations to illustrate forecast uncertainty, and present findings to cross-functional teams to inform launch strategies and initial inventory decisions.
- Implement and continuously improve methodologies for incorporating early sales signals and rapidly adjusting forecasts as new Amazon Devices enter the market.
