I take messy operational data — support tickets, network telemetry, billing exports — and turn it into something a manager can act on: clean it, model it, query it, and put the finding in front of the person who makes the call. Three projects, each one ending in a concrete recommendation.
View ProjectsA configurable engine that profiles an operational dataset and checks it against a 12-rule library spanning the six data-quality dimensions — completeness, validity, consistency, uniqueness, referential integrity, and accuracy. It produces a severity-weighted quality score, a row-level remediation file, and a formatted Excel report, and is validated to 100% detection recall against a synthetic dataset with 1,360 injected defects of known type and count.
View on GitHub →A simulated technical-support operation modelled in a normalized PostgreSQL schema and analysed with the SQL an operations analyst actually writes — CTEs, window functions, and time-series — to measure mean time to resolution, SLA compliance, technician utilisation, and time-in-status. The analysis isolates a chronic regional staffing shortfall from ticket mix and rework, and costs the fix.
View on GitHub →An end-to-end platform. A Python ETL pipeline integrates heterogeneous XML, JSON, and CSV network feeds — with per-source validation, a rejects log, and quality scoring — into a PostgreSQL star-schema warehouse, feeding a SQL analytics layer and a three-page Power BI operations dashboard. Root-cause analysis pinpoints five cell towers driving 46% of unplanned downtime and 42% of network-linked tickets, with a prioritised maintenance recommendation.
View on GitHub →