Zhenjia Zhou

Backend engineer building cloud infrastructure and agentic engineering tools.

Software Engineer at Thoughtworks. I build reliable backend systems, cloud delivery workflows, and practical AI engineering tools that survive real production pressure.

China · open to relocation abroad · zhenjiazhou0127@outlook.com · GitHub · LinkedIn · Blog

Experience

Thoughtworks

Software Engineer · Backend / DevOps Aug 2022 – Present
  • Built the async backbone of a construction-logistics delivery platform: a fault-tolerant AWS SQS + DLQ adaptor decoupling delivery changes from the core system, eventual consistency via local message tables with retries and scheduled reconciliation, and a PostgreSQL materialized-view search layer (jOOQ, wide-table design, index and thread-pool tuning) — read/write decoupled and ready for Elasticsearch migration.
  • Designed and ran the project's cloud platform from zero: multi-account Terraform/Terragrunt IaC (VPC, ALB, RDS, Cognito, WAF, Route53, ECS) with S3 + DynamoDB remote state, plus a reusable GitHub Actions library (15 actions, 8 workflows) with OIDC auth and build caching — CI job time halved (5m → 2.5m), off-peak compute cost down ~30% via ECS autoscaling, ~95% of bot traffic blocked with log-driven WAF rules.
  • Led a production go-live with service cutover and a ~50k-record data migration across Route53/ALB/ECS/RDS: authored the runbook, ran the UAT dry run, monitored the long-running migration via bastion network metrics, and completed post-migration reconciliation with zero data loss in a ~4-hour window.
  • Designed a medical-education research platform on OMOP CDM (50+ clinical tables): randomized case-assignment algorithms, Excel-driven dynamic configuration, and an LLM-as-a-judge evaluation pipeline with engineered scoring prompts — 96%+ test coverage in a data-intensive domain.
  • Drove the team's AI-native delivery transformation through harness engineering: standardized prompt workflows, token-economy practices, and agents covering the full delivery lifecycle (dev, QA, infra) — ~20% efficiency gain across the team. Co-authored a post on martinfowler.com and featured in a Thoughtworks interview; led 10+ client showcases with US stakeholders; practiced strict TDD with 90%+ coverage throughout.

Elith Inc.

Backend / AI Engineer · Contract, part-time (Tokyo) 2025
  • Delivered backend and applied-AI services for a Japanese AI company alongside my full-time role: a recruitment-matching API with LLM-powered profile summarization, and an OCR processing service for enterprise clients (Python, FastAPI).

Projects

agent-patterns-lab — Offline-first reference implementations of agent design patterns: function calling, ReAct, multi-agent, eval harnesses. Runnable examples with MockLLM and tests.

share-html — Sandboxed HTML sharing service on Cloudflare Workers + Supabase: auth, expiry rules, moderation pipeline. Includes a remote MCP server for AI agents.

retypeset-odyssey — Trilingual Astro theme powering zhenjia.dev: MDX, i18n, search, translation workflow, one-line install.

craftsmanship-skills — Agent skills distilled from Clean Code and Refactoring (npx skills add lifeodyssey/craftsmanship-skills).

de-ai-slop — Trilingual (zh/en/ja) agent skill that removes AI-writing tells.

dynamic-workflow-from-scratch — From-scratch reimplementation of Claude Code dynamic workflows: vm-sandboxed JS orchestration with journaled resume.

animichi (in development) — Agentic anime-pilgrimage travel planner: multilingual queries, geo retrieval (PostGIS), route planning, Cloudflare deployment.

Writing

Selected posts from zhenjia.dev, written in Chinese with English and Japanese translations:

Skills

Kotlin / Java (Spring Boot) · Python (FastAPI, Flask) · TypeScript · PostgreSQL · AWS · Terraform / Terragrunt · GitHub Actions · Docker · TDD · DDD · Agentic engineering (Claude Code, MCP, agent harnesses, evals)

Education

MSc, Earth and Environmental Sciences — Nagoya University 2019 – 2021
BSc, Marine Resource and Environment — Ocean University of China 2015 – 2019

Languages

Chinese (native) · English (business) · Japanese (JLPT N2)