I am early in my career, and I would rather tell that story accurately than dress it up. This is the real timeline so far: what I learned in each stage, what I built, and the habits I would pass on to anyone starting out. The full list of roles is on the experience page.
2022–2023: university, then the self-taught grind
In 2022 I started a bachelor's degree in Information Technology at HUTECH. The coursework gave me the fundamentals, but the thing that made me a developer was teaching myself the web in 2023: HTML, CSS and JavaScript first, then modern frameworks.
Most of that year was the same loop: build something, break it, refactor it. It was slow and often frustrating, and it taught me the most important beginner lesson, which is that confidence comes from shipping things, not from finishing tutorials.
2024–2025: agents and automation
Next I moved into AI on my own. I spent this period researching and implementing agentic AI and workflow automation, working with Python, LangChain, the OpenAI API and Hugging Face, and studying how autonomous agents are structured, how retrieval-augmented generation works beyond the basics, and how to use LLMs inside automated pipelines.
That research is where many of the posts on this blog come from, for example The rise of AI agentic systems and The evolution of RAG.
Late 2025: building products with small teams
From October 2025 I stopped only studying and started building products with other people:
- TAIKHOANXIN, a live marketplace for licensed accounts, which I co-founded. I built the customer storefront and the admin console. It was my first real exposure to payments, order states, warranty requests and operating a product after it launches.
- Luxe Wear AI, a multi-tenant platform for building and deploying AI agents, where I owned the product and the client application.
- DevOrbit, the open-source blog platform where I practised writing and shipping in public.
2026: cloud, speech AI and community
This year widened the scope a lot:
- From April to July I was in the First Cloud AI Journey (FCAJ) bootcamp supported by AWS. Our team built DocuFlow AI, a serverless invoice-processing pipeline, and I took the project-leader and integration role.
- From May to July I was a full-stack engineer intern and ambassador at VALSEA, working on speech AI. That is where VALSEA Classroom Copilot came from, a solo build for Vietnamese–English lectures.
- From June to July I was an ambassador for GenAI Fund's Agentic AI Build Week, and since August I have been a growth specialist intern at sixonefour labs. Together they keep me close to people building with agents and to the business side of products.
And this website was rebuilt along the way: the portfolio case study tells that story.
What I would pass on
These are the lessons I am most sure about, because each one comes from something above rather than from advice I read:
- Ship something real, early. A live product with real orders teaches you things no tutorial does. TAIKHOANXIN made me care about the unglamorous parts: order statuses, background jobs, admin tools.
- Write things down. Specs, case studies and blog posts force you to turn vague understanding into something checkable. I explore that idea in what becomes valuable when execution becomes abundant.
- Learn the stack under the framework. Moving between frontend, backend, cloud and AI made it clear that frameworks change while fundamentals such as networking, data and failure handling carry over.
- Verification beats speed. AI makes it easy to produce code quickly. Knowing whether it is right is the harder skill, and the one worth practising.
- Work with other people. Most of my growth since late 2025 came from teams, mentors and communities, not from solo study.
What comes next
I want to keep building, keep writing, and keep getting closer to the problems behind the code. If you are working on something that overlaps with any of this, you can find what I am focused on right now on the now page.