Work

Projects

Data platforms, pipelines and analytics built end to end — from ingestion to the decision they support. Each case study covers the problem, the build, and what the numbers did.

Dashboard Factory (AI)

Chat builds analytics dashboards over a real grocery dataset, drawn from a 68,706-row star schema in Cloudflare D1. The model picks a metric and a dimension; deterministic code assembles the parameterised SQL and the chart.

Open the dashboard factory

Jakarta Retail Site Selection

Location intelligence scoring every H3 hexagon in DKI Jakarta for retail and F&B opportunity, built from 289,000 Overture Maps points of interest with BigQuery and dbt. Includes a live interactive map.

Read case study

Demand-Driven Replenishment Optimization

An ML demand forecast feeding safety-stock and reorder-point decisions, scored in an inventory simulation. Fill rate rose from 81% to 98% and total cost fell 88% against a naive days-of-supply policy.

Read case study

Promotion Lift and Trade Optimization

Estimating incremental promotion lift with a counterfactual baseline, controlled price elasticity, and incremental margin, then planning a promotion calendar under a fixed trade budget with mixed-integer programming.

Read case study

Retail Demand Forecasting with AI

Store-level demand forecasting for multi-branch retail: gradient-boosted daily sales prediction with walk-forward validation, an honest naive-baseline comparison, and a GenAI layer that turns forecasts into store recommendations.

Read case study

Earthquake Analytics Pipeline

End-to-end Google Cloud data platform over approximately 500,000 earthquake events (2015-2025): Terraform-provisioned infrastructure, Kestra orchestration, a partitioned and clustered BigQuery warehouse, dbt models, and an interactive Looker Studio dashboard.

Read case study