TraceSense
A supply chain intelligence pipeline that transforms raw traceability events into predictive supplier risk scores using PyTorch embeddings and counterfactual simulation.
Key Features
Neural Supplier Embeddings
Maps each supplier to an 8-dimensional latent vector learned from historical delivery data, capturing reliability patterns beyond simple averages.
Counterfactual Simulation
Runs "what-if" inference across all suppliers for a given delivery context, enabling data-driven procurement recommendations.
Feature Engineering Pipeline
Derives training signals like actual transit days, lateness, and log-normalized quantity from raw relational traceability events.
Tech Stack
PythonPyTorchPandasPostgreSQLNode.js