Case Study

MegaMart Indonesia — Senior Data Engineer Assessment

Senior Data Engineer (L3) · Insignia AWS Business Unit · February 2026

A take-home assessment spanning platform consolidation (PySpark/Python), analytical pipelines (SQL + PySpark), and engineering leadership — all set in a realistic Indonesian omnichannel retail scenario.

Q1: Platform Consolidation

Unifying 5 Snowflake silos into a unified Bronze layer

  • 13 critical problems identified (vs. 8 minimum)
  • Schema harmonization framework with 5 source-specific transformers
  • 450-line incremental CDC pipeline with watermark, DLQ, retry, metrics
  • Technical design document written for mid-level engineers
PySparkSnowflakeCDCSchema RegistryDelta Lake

Q2: Customer 360 & Inventory Pipelines

Silver/Gold layer with exact business formula compliance

  • Customer 360 pipeline: 15 exact spec formulas, handles churned customers
  • SCD Type 2 with hash-based change detection, DDL, point-in-time SQL
  • Single-pass SQL inventory health with 4 alert flags
  • Bottleneck analysis: 4h → 25min optimization roadmap
PySparkSCD Type 2SQL Window FunctionsPerformance Tuning

Q3: Engineering Standards & Team Enablement

Professionalizing a team of 8 with 40% failure rate

  • Engineering standards guide: code, pipeline, quality, operational
  • Testing strategy with pytest examples, DQ gate design
  • Alert reduction from 500/day → <20 with severity routing
  • 90-day enablement plan with week-by-week tactics
Tech LeadershipTestingObservabilityTeam Building

Technical Stack

Languages

Python, PySpark, SQL

Data Processing

Apache Spark, Snowflake, Delta Lake

Cloud

AWS (S3, DynamoDB, CloudWatch, Secrets Manager)

Formats

Parquet, Avro, JSON

Testing

pytest, PySpark test utils

Methodology

Medallion Architecture, SCD Type 2, CDC

About This Assessment

This is the Level 3 (Senior) Data Engineer take-home assessment from Insignia's AWS Business Unit. The scenario is set at MegaMart Indonesia — the country's largest omnichannel retailer with 850+ stores, 12M loyalty members, and $2.8B annual GMV. Solutions were developed using realistic data (10 CSV files, 150MB) with intentional data quality issues baked in to test production readiness.