Archil
Chachanidze
I build data pipelines that don't break when upstream systems inevitably do — late arrivals, schema drift, partial failures, silent corruption. 7+ years on GCP, 50+ production DAGs, and enough incident postmortems to stop trusting APIs blindly. I also build the tools I wish existed when testing pipelines against chaos.
- Architected pipelines processing 1–5M rows daily across multiple BUs, 99.9% uptime
- Designed modular framework supporting 50+ production DAGs, 500+ orchestrated tasks
- Built end-to-end marketing analytics integrating GA4, CM360, Adobe Analytics → BigQuery
- Reduced pipeline failures 40% via idempotent processing and automated validation
- Led 1-year migration of 100+ legacy scripts to cloud-native GCP architecture
- Backend services and data integration pipelines using Cloud Functions and Cloud Run
- Implemented CI/CD pipelines and automated testing for stable production deployments
- Automation tools, ETL pipelines, web scrapers and REST integrations for e-commerce clients
- Automated trading systems integrating multiple crypto exchange APIs with real-time processing
- Maintained Kubernetes clusters and Dockerized services
I needed realistic, broken data to test my own pipelines against — late arrivals, corrupted records, schema drift mid-stream. Nothing off the shelf did that, so I built it. Declarative YAML schemas, 20+ injectable fault types, dual-layer persistence, WebSocket streaming. It's a test harness that thinks like an adversary.
Led end-to-end redesign of analytics infrastructure serving 3 cross-functional teams. Unified fragmented data sources into automated Airflow pipelines on GCP.