A bit late, but I wanted to write this down. Some of it is for people who do not know me yet, and some is just for myself.
2025 was the year things started to click. Less time wondering where to start, more time shipping and following through.
Things I built and shipped
My team won the AWS and DIGITIMES Hackathon with a voice-driven AI chatbot for humanoid robots, built using AWS and LangChain. We also made the finals of the Taipei and New Taipei City Hackathon.
On the side, I built a badge checker tool for the Google Cloud AI Study Jam. It was small, but seeing other participants use it made the work worthwhile.
I finished a data internship at KDAN and wrote about the work in a reflection post on Substack.
In research, I continued working on single-cell Hi-C representation learning with Graph Neural Networks. The work is slow and dense, but I presented our first poster at the Multiomics and Precision Medicine Joint Conference.
Late in the year, I passed the AWS Certified AI Practitioner exam.
The rest of the projects I worked on this year are on my projects page.
Community and meetups
I attended over 20 meetups across GDG, AWS, MongoDB, and DevOps communities. The side conversations made them worthwhile. You get to hear how engineers deal with messy problems before the solutions get cleaned up for a slide deck.
Reading
I read far more blog posts than books this year. A few stood out:
- Knowing where your engineer salary comes from. Sean Goedecke breaks down why software compensation follows business value rather than code complexity.
- You Are NOT Dumb, You Just Lack the Prerequisites. Getting stuck on difficult technical material usually means you are missing prerequisite context rather than lacking ability.
- Searching for outliers. Ben Kuhn explains why finding unusual opportunities requires exploring paths that seem irrational or uninteresting from the outside.