|
Problem → Restaurant operators reconciled sales, ops, and KPI reports manually across disconnected POS exports. Built → A production-grade, metadata-driven ELT pipeline staging raw POS reports into an analytics-ready DuckDB warehouse. Impact → One automated source of truth for sales performance, ops efficiency, and KPI reporting — replacing manual reconciliation entirely. Architecture → Bronze → Silver → Gold staging · Star schema — 3 fact tables, 6 dimension tables, 14 reporting views · Containerized, CI-automated deployment · 3 interactive Power BI dashboards.
|
|
Problem → E-commerce order data lived across HTTP APIs and SQL Server with no unified, low-latency reporting layer. Built → Parameterized Azure Data Factory pipelines ingesting 100K+ order records into a Databricks-processed, Synapse-served warehouse. Impact → Reliable, low-latency business reporting at scale with fault-tolerant multi-source ingestion. Architecture → Medallion architecture on Databricks (PySpark) · Partitioned Synapse SQL views · Fault-tolerant batch ingestion into ADLS Gen2.
|
|
Problem → Retail transaction data held revenue and segmentation insight that wasn't surfaced anywhere. Built → SQL and Python-driven analysis of 2,000+ transactions feeding interactive Power BI dashboards. Impact → Revenue trends, customer segments, and discount impact surfaced across 6 categories, 5 regions, 3 channels to support growth decisions. Architecture → SQL-based ETL workflows · EDA & statistical analysis for analytics-ready datasets · KPI-tracking executive dashboards.
|
|
🎯 Building
|
📈 Snapshot
|
| Area | Focus |
|---|---|
| ☁️ Cloud Data Engineering | Azure Data Factory · Azure Databricks · ADLS Gen2 |
| 🏛️ Data Warehousing | Azure Synapse Analytics · DuckDB · Star Schema Modeling |
| ⚡ Distributed Processing | PySpark · Staged Batch Pipelines |
| 🧮 Analytics Engineering | SQL Modeling · Data Quality · Incremental Processing |
| 📊 Business Intelligence | Power BI · Tableau · DAX · KPI Reporting |
| 🔁 Pipeline Automation | Docker · GitHub Actions · CI/CD |
⏳ Stats occasionally take a few seconds to load or need a page refresh — that's GitHub's image cache, not a broken widget. If a card ever shows blank, refresh the page once.
|