Decision Sciences and Supply Chain

From the first data foundations to production ML and a five-person Decision Sciences function.

Conceptual illustration of connected retailers, forecasting, and distribution — Conceptual illustration · Jumbotail

The problem

An early-stage B2B grocery marketplace needed data and models to support decisions across a growing catalogue, stock availability, pricing, and supply-chain operations.

What I owned

I was the founding member of Decision Sciences, working solo for nearly a year before building and leading a five-person team. I also wrote production Python algorithms for the recommendation engine and bin packing.

How I approached it

Set up the initial data warehouse and visualisation tools using Amazon Redshift.
Built production recommendation algorithms and ML-driven forecasting; a separate model estimated sales lost during stockouts and fed demand forecasts.
Built systems across pricing and supply-chain optimisation, while growing the function and supporting prioritisation and decision-making across the company.