HZHuzhi (James) Zhao
Open to new opportunities

Senior Backend Engineer

expanding into Data Engineering

Hi, I'm James — a backend engineer with 9 years of experience building distributed systems where scale, performance, and data reliability matter.

Now based in Winnipeg, I'm extending that experience into Data Engineering and Applied AI through post-degree study and hands-on platform projects.

See my workLet's Connect
James@winnipeg:~
$ ▌
PROJECTS

Projects I've Shipped.

A couple of projects showing how I take systems from raw data to production.

UOIP 示意图：311 报告、排班表与气象数据合流，按降雪事件切分，输出各分区的运行负载排名
Day of Data 2026
A self-hosted lakehouse ingesting City of Winnipeg open data through Bronze, Silver and Gold layers into a per-zone winter operational load score.
Operational Data Platforms
2026

Urban Operations Intelligence Platform

How does a city know where to send its plows after a snowfall?

AirflowSparkMinIOTrinoLakehouse
ToucanShelf 示意图：左侧层级目录树与 gallery 视图，右侧一篇文档内的日历、表格、draw.io 与加密块
A fork that adds folder trees, gallery views, encrypted and interactive blocks, git round-trip sync and public publishing — one Go binary, one SQLite file.
Applied AI2026

ToucanShelf — Self-Hosted Knowledge Base

What turns a stream of notes into a knowledge base a team can live in?

RAGMCPGolangReactSQLite
Factory MES 产线示意图：24 道工序串成一条主线，Web 看板与安卓工位端分支接入
Production
Closed Source
A production-grade manufacturing execution system built solo — Go services, RBAC, operational dashboards and Android shop-floor apps.
Industrial Systems
2025

Factory MES & Analytics Platform

How do you get 24 production stages onto one system?

GolangVue.jsMySQLAndroidRBAC
图标标注管线示意图：爬取的图片经 CNN 过滤，由 CLIP 与 BLIP 描述，再以自身标签重命名
A fine-tuned CNN filters scraped images, CLIP and BLIP describe them, and each icon is renamed after its own tags, making filename search semantic.
Applied AI2025

Vision-Model Icon Annotation Pipeline

How do you make 1,783 icons searchable without writing a single keyword?

CLIPBLIPGemini APIResNet50V2Computer Vision
EXPERIENCE

Where I've Worked.

Nine years of backend engineering across e-commerce, SaaS, and manufacturing — now moving into data and AI.

Education

  • University of WinnipegUniversity of Winnipeg (PACE)Jan 2026 — Dec 2027Winnipeg, MB, Canada

    Artificial Intelligence Post-Degree Diploma · Business Analysis & Transformation Post-Degree Program

    End-to-end delivery: business problem framing, data pipelines, and ML engineering.

  • Henan University of EngineeringHenan University of EngineeringSep 2010 — Jun 2014Henan, China

    Bachelor of Engineering, Mechanical Design, Manufacturing and Automation

2024 — 25

Shanghai ZHONGYOU TIPO Steel Pipe

Shanghai ZHONGYOU TIPO Steel Pipe Co., Ltd.

Jul 2024 — Feb 2025Remote, from Hangzhou, China

Sole developer on a production MES covering 24 manufacturing stages.

GolangVue.jsMySQLAndroidRBAC

Full-Stack Developer (Contract) — Manufacturing Execution System

Steel pipe manufacturer. Sole developer on a greenfield MES, end to end.

  • Designed, built, and deployed a production-grade MES from scratch as the sole developer — architecture, backend, frontend, database, mobile, deployment, and production support.
  • Modelled complex manufacturing workflows across 24 production stages, translating shop-floor processes into scalable software.
  • Architected backend services in Go with relational data models covering production tracking, quality control, inventory, and manufacturing traceability.
  • Designed RBAC authorization and workshop-based data partitioning to enforce operational security and process ownership.
  • Delivered web admin portals, operational dashboards, reporting, and Android shop-floor apps, with real-time KPI monitoring, defect analysis, and equipment status reporting.
See it in Projects
GolangVue.jsMySQLAndroidRBAC

2021 — 23

Weimob

Weimob

May 2021 — Nov 2023Hangzhou, China

Owned the Elasticsearch write path, migration tooling, and data-reliability controls for a search and promotion platform serving 20,000+ merchants — upstream quota 3,000 → 10,000 QPS, publish latency 30+ → <10 min.

JavaElasticsearchKafkaRedisMySQLDubbo

Senior Java Software Engineer

Enterprise SaaS commerce platform. Owned search and promotion infrastructure for 20,000+ merchants.

Selected engineering work

Scaled million-message promotion updates

A large campaign could fan out into millions of activity-by-store updates, while an upstream product service initially limited the pipeline to 3,000 QPS. I split update events by change type, isolated oversized workloads into dedicated processing paths, and negotiated a quota increase to 10,000 QPS after reducing unnecessary upstream reads. End-to-end publish latency fell from more than 30 minutes to under 10.

Rebuilt the index topology during a live migration

During a seven-month backfill, a data-heavy account representing more than a third of indexed data pushed one node past Lucene's 2.1-billion-document ceiling and stopped writes. I restored writes by separating large accounts into a dedicated cluster, then redesigned routing and query paths while the legacy system continued serving customers. The platform grew from roughly 65M to 90M indexed product records across 32 data nodes; hot/cold separation, filter-only queries, and index sorting kept query latency stable without adding nodes.

Turned migration jobs into an operations platform

Backfill, reindex, and repair jobs ran for months against the same clusters and upstream quota as live traffic. I built a distributed task platform with five priority queues, backpressure, lock-based idempotency, emergency stops, alert controls, and dead-letter replay. It remained in use after the migration as the team's shared platform for backfill, repair, and index operations.

Made silent data divergence detectable

MySQL and Elasticsearch could drift apart without warning, so the first signal sometimes came from a customer. I built a configuration-driven, multi-tenant reconciliation service that compared data by field, scope, and document count, then supported alerts, retries, repair, and an audit trail. The work turned silent divergence into a detectable and recoverable condition, and was later incorporated into a company-wide reconciliation platform used by multiple internal teams.

JavaElasticsearchKafkaRedisMySQLDubbo

2019 — 21

DaSouChe

DaSouChe (NASDAQ: DSC)

Mar 2019 — May 2021Hangzhou, China

Led the Vehicle Product Center platform serving 40,000+ automotive businesses.

JavaRule EngineElasticsearchMulti-tenant SaaS

Senior Java Software Engineer

Used-car trading SaaS platform for automotive dealers across China; listed on NASDAQ in 2026.

  • Led development of the Vehicle Product Center platform serving 40,000+ automotive businesses.
  • Designed configurable workflow and business-rule automation inspired by Salesforce Flow, letting non-technical users define event-driven processes without code changes.
  • Developed a metadata-driven rule engine on Elasticsearch, Redis, MySQL, and Painless scripting.
  • Built scalable backend services and APIs supporting product lifecycle management and tenant-specific customization.
JavaRule EngineElasticsearchMulti-tenant SaaS

Additional experience

  • Career break — relocation to CanadaMar 2025 — Dec 2025Hangzhou, China

    Self-directed study — IBM AI Engineering Professional Certificate.

  • Java DeveloperHangzhou Tanhua E-commerceDec 2023 — Jun 2024Hangzhou, China

    Supply-chain and ERP maintenance, standalone-site integration, marketing automation.

  • Junior DeveloperZhejiang Sendinfo TechnologyMar 2017 — Mar 2019Hangzhou, China

    B2C ticketing SaaS — ordering, refunds, validation, payments — and a group-tour ERP.

  • Junior DeveloperZhejiang Yonyou SoftwareJan 2016 — Mar 2017Hangzhou, China

    Points-based loyalty malls, and mobile field-data apps for infrastructure contractors.

CONTACT

Let's Connect.

Open to backend, data engineering, and AI platform roles in Manitoba.

Based in Winnipeg · Open across Manitoba
Email[email protected]
LinkedInlinkedin.com/in/huzhi
GitHubgithub.com/huzhi-zhao
LocationWinnipeg, MB, Canada