2021–2022

Hypertrade — Sales & Promotion Forecasting at Scale

PythonTensorFlowScikit-learnGoogle BigQueryFlask

Designed, trained, and deployed ML models for sales and promotion forecasting for international retailers, achieving 79% average prediction accuracy.

Data Scientist

Challenge

Retailers needed accurate sales and promotion forecasts at scale to optimize inventory and promotions. Existing methods were not sufficiently accurate or automated.

Implementation

Designed, developed, trained, and deployed machine learning models for Sales and Promotion Forecasting. Used Google BigQuery, Data Studio, and Compute Engine. Served models via Flask APIs on Google App Engine. Managed MLOps with MLFlow and automation with Ansible and AWX.

Selected stack

PythonTensorFlowScikit-learnGoogle BigQueryFlaskMLFlowAnsibleAWX

Outcome

79% average prediction accuracy for sales and promotion forecasting; production pipelines and MLOps in place for international retailers.