Mobile networks
I started in the county sales department as assistant to its sales operations supervisor. Six months later I moved into base-station maintenance and then network optimisation as a communications engineer, earning the engineer qualification and an equipment-maintenance certification. The work was demanding, and I reasoned that if every job would be hard, I might as well do the one I enjoyed. I had studied computer science and liked it, but IT roles were scarce in the company branches I could reach, so in 2008 I moved to Beijing to work in enterprise IT.
Enterprise systems, backup and storage
At Beijing Kunlun Unicom, an enterprise software reseller, I began as a post-sales engineer. My main focus became Symantec (Veritas) backup solutions, Backup Exec and NetBackup, which I deployed and supported across Windows, Linux and UNIX, alongside some Microsoft Exchange work. Backup work taught me to understand the systems being protected, which brought me into AIX, Solaris, Oracle and DB2. I earned Symantec Backup Exec and storage certifications.
We looked after whole product lines, so the same engineer covered pre-sales and post-sales, and I soon took on pre-sales as well: understanding what a customer ran, writing the proposal and presenting it. If the customer accepted the proposal, they bought the products from us together with installation, configuration and maintenance services, which I then delivered. Because I implemented what I proposed, my designs had to fit the customer’s real environment: technically feasible, well reasoned and able to meet their goals. That focus on practical, achievable solutions carried into my later consulting work.
I then did similar pre-sales and implementation work at another company, followed by a project-coordination role. By then I wanted to focus on systems and database work. In backup projects I had noticed that the experienced engineers all knew Oracle well, so I studied Oracle for about three months and earned the Oracle Database 11g Administrator Certified Professional (OCP) in October 2013.
From operations to containers
At Zhongrun Sifang I worked as a system engineer, covering both systems and operations, including application releases. One of the company’s major projects was a national tax e-invoice cloud platform, and I worked on it. For the application release platform built for Jiangsu State Tax, I prepared and maintained the underlying environment: servers, operating systems, and network, storage and database requests. The development team built the release functions, and an OpenStack vendor designed and deployed the underlying OpenStack environment. Seeing how a release platform depends on every layer beneath it made me want to be involved earlier, in the design.
Alongside operations I studied technologies the company might use: OpenStack first, then Hadoop and HBase. When the company chose to buy mature OpenStack products rather than build its own, I wanted more say in technology choices and platform work, so I turned to containers. Later in this role I wrote my own Ansible scripts to build Kubernetes clusters, so I could test how applications were deployed and scheduled. The company used containers much like small virtual machines, with many manual release steps, so I proposed changing that and introducing Kubernetes. The idea was not adopted, but it prepared me for my next role.
A first client container platform
On Huawei’s HiCloud project I was the only Kubernetes specialist. The client was a small team with a limited budget that wanted a container platform it could deploy, manage and maintain itself. I designed the container cluster solution, researched Kubernetes scheduling, built the base cluster, automated the platform deployment with Ansible, and worked with two or three developers to deliver the features the client needed. The client team had no Kubernetes experience, so helping them build that capability was part of the job. That included a young engineer the client had brought in from an outsourcing firm to work on Kubernetes. He started with only container basics; I mentored him throughout the project, and by the end he knew Kubernetes well.
It was my first time applying Kubernetes for a real client. I had built clusters to learn; here every feature had to meet an actual requirement, and my understanding of Kubernetes deepened as we built them one by one.
From releases to continuous delivery
At Shenzhen Huayun Zhongsheng, I advised a platform-development team on Docker and Kubernetes architecture. I then joined Shenzhen Xinzhen, a financial platform company, as a DevOps engineer. Its release logic lived in a long, complex Bash script. I read it section by section, then asked the engineer who wrote it to confirm my understanding of each step. Once I understood the release logic, my daily focus moved to how software is continuously delivered: code in GitHub, Jenkins pipelines, scripts and container releases. This was my first formal DevOps role.
From specialist to delivery and consulting responsibility
I joined Thoughtworks as a Senior Consultant with deep Kubernetes experience. Agile delivery was new to me, and I learned it on client projects. My first projects meant building pipelines and explaining DevOps practices to a bank team, then coaching developers from a vendor the client had brought in, so they could take on Kubernetes support for a second, larger Huawei container platform. Between projects I studied agile practice, read The Phoenix Project, wrote internal articles and gave knowledge-sharing sessions on Istio and Go. I also earned the Certified Kubernetes Administrator certification, CKA (CNCF, 2019).
China Merchants Bank, 2018–2020
This container platform engagement is where my role changed most.
- Phase 1. I led the initial assessment and the design of a hybrid container and VM platform, working with the bank’s architect on business requirements and legacy constraints. One example: the bank’s database access rules identified each application by a fixed source IP and port. I designed a host-level addressing approach, with preallocated address ranges and a set number of containers per host. I weighed the scheduling trade-offs and chose a design that met the rule; it passed the bank’s security review.
- Phase 2. As the platform expanded to multiple OpenShift clusters and added VM scheduling, I led the design for this phase. I was also the infrastructure, DevOps and container-platform consultant throughout, guiding the development team on infrastructure, deployment and platform issues, and I started business analysis for the platform’s operations features.
- Phase 3. By then the platform’s core features were mature, and the work turned to how it should evolve. Thoughtworks brought in two of its most senior infrastructure Tech Principals to work out the evolution direction and plan with me. The design work was shared: the Tech Principals and a senior data engineer led some parts, and I led others. Alongside my continuing role as the container-platform consultant, I took on more of the requirements analysis: analysing the bank’s operational scenarios, writing user stories and acceptance criteria, and explaining them to developers and data engineers. I also handled part of the project and team management. We delivered the phase through the COVID period, working remotely.
- Phase 4. I contributed to early design before moving to other projects.
The bank operated the platform, which went on to support more than 1,000 enterprise applications. Much of a release’s time went into the bank’s own processes and approvals; the platform automated as much of the rest as it could. Once it was in place, release cycles fell from about two weeks to two or three days.
Other engagements
- Costa, 2019. As a DevOps engineer, I developed the application monitoring and log-alerting workflow connecting Graylog, Alertmanager and Slack. Each alert carried its environment and a link to the matching logs, and alerts could be silenced for a set time. A separate scheduled check tested the homepage and reported the result to Slack.
- Daimler, 2020–2021. As DevOps Team Lead, I managed the environments for two development teams and led one other DevOps engineer. I migrated the gopass-based credential workflow to a highly available HashiCorp Vault on OpenShift, cutting password synchronisation for 200+ users from over an hour to seconds. I also implemented Pipeline-as-Code in GoCD and used Nexus as the artifact repository.
- ICBC Zhuhai, 2021. As a DevOps consultant for about three months, I analysed the delivery problems the bank had raised, reviewed its full release path, and proposed solutions for deployment standardisation, delivery metrics, test-environment database changes and traffic switching. Over several rounds of review with the bank’s teams, I refined them to fit its release process and existing systems, ready for its own teams to implement.
- Authority Brands, 2021. As SRE support on its AWS platform, I refactored Terraform modules and built service modules for a recovery environment. With a colleague, I assessed the platform’s reliability and security, including monitoring, tenant isolation and recovery readiness, and prioritised our recommendations by risk and implementation effort.
- SadaPay, 2021–2022. As a cloud infrastructure engineer, I provisioned AWS infrastructure with Terraform, developed modules to support disaster recovery, and supported EKS and the team’s existing Argo CD and GitHub Actions delivery, on a fintech platform serving more than a million active users across 20+ microservices.
- Health Promotion Board, Singapore, 2022–2024. As a DevOps engineer on the engagement, I supported the development teams day to day on Azure and AKS. To meet mandatory government security requirements, I led and completed the environment migration that consolidated the project’s environments and their Terraform and deployment runner VMs in a government-managed Azure subscription, with another DevOps engineer supporting me. I guided him throughout the project. I rebuilt the pipelines in GitLab CI/CD and prepared the runners with Ansible. When a required security scan was blocked by missing access, I set out what was needed and worked with the product owner and the client’s support team to secure it, and we then passed the required checks. I also prepared materials on our Terraform plan and Azure resource-import approach and gave knowledge-sharing sessions on them for client staff and teammates.
All client engagements, with their case studies, are listed under Enterprise experience.
A new start in New Zealand
At Thoughtworks I enjoyed working in an international company: on projects for clients in different countries, with colleagues from many backgrounds. That made me want to continue my career overseas. A few months after leaving in July 2024, my family and I decided to move to New Zealand, where I saw a realistic path to building a long-term career and a life. I chose to study here first, then continue my infrastructure and DevOps career.
- I prepared for and passed PTE Academic in March 2025 to meet the master’s entry requirement, and moved to Auckland the same month.
- I completed a Master of Software Engineering (NZQF Level 9) at Yoobee, conferred in July 2026. My capstone compared single-pass, chain-of-thought and two-role generator–reviewer pipelines for summarising enterprise IT tickets, using the Gemini API and a locally run Llama 3 8B.
- For Story Filler, I moved the Django backend to Google Cloud Run and managed the cloud infrastructure with Terraform, and wrote up the approach on Medium.
- After years of using Terraform on client projects, I took HashiCorp Certified: Terraform Associate (004) to consolidate the fundamentals, and earned it in October 2026.
- I am now working through AWS Solutions Architect – Associate topic by topic, to refresh and organise the fundamentals behind the AWS work I have done in projects.
- I use AI coding tools such as Codex, Claude Code and Antigravity for planning, implementation and debugging, and check their output against the source, tests and running behaviour.
- I turned my main projects into the case studies on this site, checking each one for what I did and what the team or client did.
What changed
Over these years, my work moved from proposing and delivering backup solutions for customers, to designing container platforms for clients, to leading delivery with development and client teams.
Most of what I learned along the way, I went looking for myself, and I still do.