About
I’m Jiajun (Eddy) Huang, an AI/ML software engineer studying Artificial Intelligence at Northeastern Silicon Valley after Computer Science at UC Davis. My work sits between model behavior, product judgment, and the quiet engineering that makes both survive contact with real users.
Now
M.S. AI · Northeastern
Based in
San Jose / Bay Area
Looking for
Fall 2026 internship
01 / 10
Operating notes
I like mapping the boring parts first: data shape, user loop, failure modes, and what success should actually look like.
For AI systems, evals are not paperwork. They are how you keep the product honest after the demo works once.
I enjoy polished interfaces, but I trust shipped behavior more than aesthetic confidence. The system has to hold.
The longer version
A short route through the instinct, the training, and the kind of AI work I want to do next.
01
China · before the tools
I grew up noticing how much everyday work is a loop: the same steps, in the same order, handed from person to person. That bothered me before I had the vocabulary for software, and it still shapes how I think about engineering.
“Understand the loop before trying to automate it.”
02
UC Davis · 2020 — 2024
At UC Davis, Computer Science turned that early instinct into craft. I spent a lot of time with algorithms, machine learning, and the less glamorous parts of making software actually run. The through-line was simple: I liked code most when it helped a system decide.
“The best code is usually a decision made legible.”
03
Northeastern SV · 2025 — 2027
Now I am pursuing an M.S. in Artificial Intelligence at Northeastern Silicon Valley. I am especially drawn to retrieval, tool-use, agentic workflows, and the small reliability details that decide whether an AI product feels clever or actually useful.
“I care about systems that can act, recover, and be evaluated.”
04
Fall 2026 · Bay Area
For Fall 2026, I am looking for SWE, ML, or AI internship work close to real users. I want the problems to be hard, the feedback loop to be short, and the team to care about latency, evaluation, and product judgment as much as model capability.
“A useful system beats a flashy demo every time.”
