Aaron
Hey, I'm Aaron — a software engineer with a machine learning research background and hands-on experience across full-stack product development, cloud infrastructure, and DevOps. I published an IEEE paper at Academia Sinica before moving into product engineering and platform work. Outside of work I'm an avid runner and triathlete, always pushing myself in both tech and endurance sports.
Latest from the Blog
Thoughts on backend, infra, and beyond.

RabbitMQ 中 Message Persistence and Durability
如何透過 Durable Queue、Persistent Message 與 Publisher Confirms 三者搭配,確保訊息在 Broker 重啟後不遺失

什麼是 RabbitMQ
介紹 Message Queue 的概念,以及 RabbitMQ 的核心組成元件:Producer、Broker(Exchange、Binding、Queue)

Deep Dive Into AI Agent
隨著 LLM 越來越聰明,所支援的 Context Window 越來越大,相關的應用開始從單次的 LLM chat 到現在每間公司都在嘗試產品化的 AI Agent。 Agent 本身到底是什麼呢?
Skills & Expertise
Technologies and tools I work with day to day.
Backend
Frontend
Cloud & DevOps
Observability & ML
Changelog from my journey
From machine learning research and full-stack product development to leading GCP cloud infrastructure and building DevOps & observability from the ground up — here's a timeline of how I got here.
- Introduced Terraform to manage GCP infrastructure, covering network design, security hardening, and egress cost optimization
- Designed the full GitLab CI/CD pipeline integrating lint, test, SonarQube, and Artifact Registry, with Helm-based deployments to GKE supporting multi-environment and rollback
- Planned and implemented an on-premise API Gateway with Keepalived for an HA architecture
- Built Prometheus + Grafana + Loki observability stack from scratch, designing metrics, dashboards, and alerting
- Redesigned the centralized log sink and retention policy, reducing logging cost by ~60%
- Lead developer of an AI Agent product — responsible for system architecture, core development, and AI observability
- Deployed n8n and introduced automated workflows to streamline cross-team collaboration
- Optimized CI pipeline, reducing CI time by 33%
- Planned and executed MySQL upgrade from 5.7 to 8.0
- Upgraded two products from Ruby 2.7 to 3.3 and Rails 6.1 to 7.1
- Full-stack product development — owned feature architecture design and backend implementation
- Built AI project backend architecture, automated testing, and MySQL / ElasticSearch backup mechanisms
- Implemented and analyzed various meta-learning approaches for few-shot multi-class training.
- Published an international paper at IEEE SMC.
- Implemented various data preprocessing methods
- Implemented and studied multiple reinforcement learning frameworks and related research papers