← Portfolio
Proprietary

Data and ML platform for an optical distributor

Industry
Optical retail
Type of work
Data engineering & ML
Period
2024 – 2026
With
Owl Tech Labs ↗ (founding partner)

Confidential project: code and data are not public.

Architecture

ERP billing SharePoint catalogs Bronze raw copy Silver customers, products,suppliers, invoices Gold (dbt) sales, equivalences,recommendations Dashboards sales team ML LLM attributes +classifier · MLflow Dagster orchestrates everything · Kubernetes · Postgres · IaC · CI/CD

Goal

Unify billing and catalogs from several suppliers into one data model, and use it to recommend the in-house equivalent for each third-party sale.

What I did

  • Designed the ELT pipeline in Dagster and dbt, with bronze, silver and gold layers.
  • Wrote the normalization of customers, products, suppliers and invoices, which arrived in different formats.
  • Built the ML side: LLM-based product attribute extraction and a supervised classifier, tracked in MLflow.
  • Built the product-equivalence chain that feeds the sales dashboards.

Stack

DagsterdbtPythonLLMsMLflowKubernetes