Dewei Zhai 
Dewei Zhai
Freelance data engineer & AI practitioner
I'm a freelance data engineer based in the Netherlands. 8+ years building TB-to-PB data lakehouses across AWS, Azure, and GCP for these clients:
PVH Corp 2020–2022 · 2024–2025 · Lead Data Engineer (AWS + Azure Datalake) 500+TB data lake on AWS, with Azure-side integrations. 90+ sources, 1000+ datasets. Led ETL, platform architecture, CI/CD, data quality. Real-time GDPR (de)anonymization (10× cheaper). 500+ TB 
VodafoneZiggo 2022–2024 · Data Engineer & Infra Admin 1PB+ datalake. Oracle → Snowflake migration, CDC ingestion (DMS), AWS IaC (CDK + Terraform). New CI/CD saved the team 60+ hours. 1 PB+ 
FedEx Digital 2018–2020 · Data DevOps Engineer ETL pipelines on AWS & GCP, productized data-science models, Kinesis-based streaming. 200 TB 
KPN 2016–2018 · Data Engineer / Hadoop Admin Hadoop administration, Hive/Spark ETL, automation with Ansible & Jenkins. 42 TB 
ABN AMRO 2018 · Data Engineer Hive/Spark ETL, contributed to enterprise data lake build. 100 TB For the past 2 years, I've gone deep on AI-augmented development — turning fuzzy problems into ones AI can reliably solve. Daily tools:
Cursor 
Codex 
Claude Code Currently the data lead at an AI-native startup in a heavily regulated domain:
Enyquant 2026 — present · Platform Architect & end-to-end Data Engineer Sole hands-on architect for a dual-region Lakehouse (Azure EU + Alibaba Cloud China) at an energy-trading, AI-first startup. Full ownership of platform, pipelines, IaC, CI/CD, IAM, and cross-region architecture. ~3× delivery throughput via AI-augmented practice. Recent writing
All posts →-
Secrets in the Keychain, not in your .env
A 230-line Bash wrapper around the macOS Keychain that lets me inject API keys into one command at a time — without ever exporting them to my shell or committing them to .env.
2026-05-15
-
Fix the contract, not the dashboards
At PVH the ticket said 'fix the dashboards.' The dashboards weren't the bug — the contract between two teams was.
2026-05-13
-
CLI, not MCP
Swapped a Python Outlook MCP for a 250-line Rust CLI. The email body the LLM sees dropped from 6,200 to ~500 bytes, and 700 tokens of tool schemas left the context.
2026-05-13
Have a talk?
Synchronous
Talk to my agent — it knows my track record and will tell you whether I'm a fit before you spend time writing.
Asynchronous
Leave a brief or email me directly. I read everything; I reply to what I can help with.