Production data work, measured.

Founding/early data engineer building the ingestion and structuring layer and retrieval stack for a legal research product over Pakistan's statutory and case-law corpus.

• 01Built high-throughput ingestion pipelines for Pakistan's legal corpus — fragmented PDFs and HTML to structured, queryable records, drawing from ~15 sources.
• 02Designed the retrieval layer with full-text + dense-vector search, consolidating a fragmented Turso/Qdrant/MeiliSearch stack onto a PostgreSQL-centric architecture.
• 03Improved section-level statute retrieval for AI systems to ~80% accuracy on internal evaluation queries, where weak document-level FTS had been failing.
• 04Led the data function as the responsible owner for pipelines, documentation, and handoffs across 1 full-time engineer and interns.

High-volume advertising data pipelines with a focus on operational reliability: automated monitoring, consistent checks, and fast failure detection.

• 01Automated delivery-failure detection and Salesforce case creation, cutting manual monitoring effort by over 90%.
• 02Built and maintained Airflow / Google Composer workflows supporting delivery to 50–60 downstream destinations (DV360, Trade Desk, Yahoo).
• 03Implemented recurring reporting and file-runner workflows (BigQuery checks, Compute Engine jobs, SFTP delivery) handling ~100 files/day.