Led a 3–4 person engineering team across data and service-platform projects at AmazingTalker.
Given
Business data, advertising data and in-house tracking lived in separate systems. Marketing and product teams depended on engineering work to prepare reports.
Warehouse
Built a BigQuery warehouse with Datastream CDC, Airflow and dbt, plus Metabase for self-service reporting. It unified business data with GA, Meta, Google Ads, TikTok and in-house tracking.
Platform
Built an Nx monorepo with a service generator, shared libraries and CI/CD that built and deployed only affected services to EKS. Product teams adopted it to decouple services from the main app.
30% → 5%of sprint time spent on reporting, approximately
3 daysnew data API delivery, down from about a week
The client needed staff to ask natural-language questions over private documents, without sensitive content leaking between users, and without manual data prep for every new file.
Method
Built a retrieval-augmented-generation pipeline: automated ingestion, chunking and embedding into a vector store, with an LLM answering only from retrieved, permission-scoped context.
Decision
Made per-user data isolation a first-class part of the design rather than an afterthought, so security held up even as the document set and user base grew.
Full batch recomputation delayed ranking updates: new sales took about 20 minutes to appear in teacher rankings.
Method
Rebuilt ranking with Elasticsearch scoring and Redis sorted sets per locale. Event-driven updates recalculated individual teachers instead of the full batch.
Decision
Updated each teacher as their data changed, keeping rankings current without waiting for the next full recomputation.
Video
Launched teacher short videos using S3, SQS fan-out, ffmpeg on Lambda and HLS via CloudFront. Playback tracking fed teacher ranking and video recommendations.
20 → 2 minfor new sales to appear in teacher rankings
−60%ranking API latency, from 500 ms to 200 ms
1,000+video transcoding jobs per day
SubjectAgents, cloud and automation
Sheet3 of 6
Selected work
Applying production engineering to agents, cloud systems and developer automation.
A CLI that runs an AI agent against a GitHub or Linear issue inside a disposable Docker sandbox. It implements the change, runs the project's tests, scans the diff for leaked secrets, does an independent read-only review, and opens a pull request only once every gate passes: issue to code to review to PR, fully automated. The direction is "loop engineering": a fast-to-deploy workflow you drop into any repo so AI can develop, review and raise PRs alongside your CI/CD.
Fig. 2A run from issue #72: preflight, setup, implement, verify, review, then the pull request.
A cloud-native conversational assistant for a user's own Strobe posts and photos. It combines Amazon Bedrock, grounded vector retrieval, a separate MCP service, asynchronous media classification and scheduled recaps, with Cognito identity and explicit permission for mutations.
Fig. 3Asked for dog photos, the assistant returns matching images and a story grounded in them.
An agentic knowledge-management tool that turns raw, messy notes into structured knowledge, generating update proposals, tracking recurring mistakes, creating review tasks and building readiness maps. A built-in RAG layer lets users ask questions in natural language and instantly surface the right note or past memory from their own knowledge base. Built on the OpenAI Agents SDK with a clean-architecture Express backend and a React + Vite client.
Fig. 4An update proposal with its source evidence, waiting for a human to apply or reject it.
No. 4Industry project
AI Meeting Management Platform
An on-premise government proof of concept built as a QUT industry project, turning recordings into draft minutes and action items.
Design choice
Keep audio and model processing on-premise; use durable asynchronous workers so long recordings do not block the API.
Built as a government proof of concept for the Department of Science and Technology (DOST), Philippines, through a QUT industry project, with me as the primary developer in a Brisbane-based team of six. Whisper transcribes recordings and a local Llama 3 model drafts minutes and action items for staff to review on the client's infrastructure. The Express/PostgreSQL backend handles scheduling, search, file storage and role-based access, with CI/CD and Jest tests. A transactional outbox and Redis Streams provide durable job dispatch across Node.js and Python workers; SSE reports progress without blocking the API.
Fig. 5Action items extracted from meetings, tracked across To-do, In progress and Done.
SubjectRoles, most recent first
Sheet4 of 6
Experience
Team leadership, backend architecture and end-to-end delivery across production systems and client projects.
Feb 2026 – Present
Full Stack Engineer
QUT Industry Project · Brisbane
Primary developer in a six-person industry team, building an on-premise government AI proof of concept with an Express/PostgreSQL backend and durable Node.js/Python workers.
Owned the NestJS/Prisma/PostgreSQL backend of a B2B garment-manufacturing MVP, including Stripe payments and an artwork validation and vectorisation pipeline.
Jul 2023 – Oct 2025
Full Stack Engineer (Part-time)
Freelance · Taiwan
Delivered permission-scoped RAG, layout-preserving translation for 100+ page financial reports, and an iGaming CRM used by dozens of operations staff.
Led a 3–4 person team across data, ranking and service-platform work. Built engineering guidelines and an onboarding project used by ~10 engineers across teams, typically shipping to production within 3 days.
Removed Redis KEYS scans and queued settlement with RabbitMQ, cutting API time from 400 to 50 ms. Built an idempotent wallet load-tested to 2,000+ QPS and cut deployments from 30 to 5 min.
Aug 2018 – Aug 2019
Backend Engineer
Pentium Network Technology · Taiwan
Integrated CDN and FQDN resources across AWS, Alibaba Cloud, Tencent Cloud and GCP; built chatbot and workflow-automation systems with Kafka for real-time event processing.
Apr 2017 – Jun 2018
Backend Engineer
Lingtelli Co., Ltd · Taiwan
Built the company's chatbot product for Facebook and LINE and its natural-language-understanding engine, with auto-scaling and CloudWatch monitoring.
One line per role. The full bullet points are in the résumé (PDF).
SubjectBackground and skills
Sheet5 of 6
The engineer behind the work.
I'm a full stack and backend engineer with 6+ years building systems, from chatbot and NLU products to distributed services, data platforms and LLM-powered applications. At AmazingTalker I led a 3–4 person engineering team; at Sourect I owned backend architecture for a B2B MVP. I also spent two years freelancing part-time, designing and delivering client systems end to end.
What I care about most is the part users actually feel: speed, reliability, and whether things still hold up as they grow. I like taking something slow, fragile or messy and turning it into something fast, observable and easy for the next engineer to pick up, and I try to back decisions with real numbers rather than opinions.
Right now I'm completing a Master of IT at QUT in Brisbane, and most of my own time goes into agentic AI, building things like an autonomous issue-to-PR coding agent and a tool that compiles raw notes into structured knowledge. I'm looking for backend or AI engineering roles where I can keep working where solid systems meet applied AI.
On the practical side: I'm based in Brisbane on a student visa with work rights now, and on finishing my Master I move to a Temporary Graduate visa (subclass 485) with full, unrestricted work rights, so no employer sponsorship is needed in the near term, with a longer-term view to permanent residence.
Languages & web
Go
Python
TypeScript / Node.js
Ruby
C#
NestJS / Express
FastAPI
React / Tailwind CSS
REST / GraphQL
Data & events
PostgreSQL / MySQL
Redis / Elasticsearch
BigQuery
dbt / Apache Airflow
CDC
RabbitMQ / Kafka
Event-driven architecture
Cloud & DevOps
AWS / GCP
Docker
Kubernetes (EKS)
GitHub Actions CI/CD
New Relic
Nx
AI & tools
RAG / Vector databases
Llama 3 / Whisper
Amazon Bedrock
Agentic workflows
MCP / ACP
Claude Code
Education
Qualification
Institution
Dates
Master of Information Technology
Queensland University of Technology, Brisbane
Feb 2025 – Nov 2026 (expected)
Bachelor of Computer Science
National Taiwan University of Science and Technology, Taiwan
Open to senior backend and applied AI roles. Based in Brisbane, available for local and remote opportunities.
Student visa with work rights now. Full work rights on a Temporary Graduate visa (subclass 485) after my Master in Nov 2026, so no sponsorship is needed.
