About

I’m a Computer Science student at the University of Michigan, drawn to the layer where systems engineering meets intelligent software — distributed systems, backend infrastructure, and the machine learning that increasingly runs on top of both.

I tend to think of what I do as building in the fuller sense of the word: not a single act of engineering, but the repeated cycle of planning, executing, reflecting, and refining. It’s why I care about correctness before speed, depth over breadth, and measuring things rather than assuming them. The projects I’m proudest of came out of that loop, not out of a first draft.

What I build

CLAI began as a small frustration. Command-line tools are powerful and almost aggressively unfriendly — you look up flags, memorize syntax, and hope the machine in front of you behaves like the last one. So I embedded a language model into the terminal: you describe what you want in plain English, and it translates that into the precise command. The real lesson arrived when it nearly wiped one of my directories. The problem was never translation — it was trust. Irreversible actions now confirm before they run. CLAI went on to win first place in the Google track at MHacks.

Other work runs deeper into systems. I’m implementing the Raft consensus algorithm from scratch in Go, with a fault-tolerant replicated key-value store layered on top — the kind of project where “it passed once” means nothing, and only a long run of consecutive passes under injected partitions and crashes counts as done. Before that, I spent a summer on independent research in multi-agent reinforcement learning, designing and benchmarking a soft actor-critic variant for competitive environments and writing up the results as a paper.

My aim is to keep a foot in both worlds — rigorous engineering and modern AI — whether that leads toward building infrastructure at scale or research at a lab.

Beyond the terminal

Away from the keyboard I play soccer, which is quite a pivot but a way to break from academics and alike: technique, repetition, and the easy company of teammates.

I’ll also admit to one long-running obsession: the meta of Honor of Kings. Most people see a mobile MOBA; I see an ecosystem of real-time decision-making, where a half-second shift in a crowd-control effect can ripple through a team fight and retire an entire strategy, or a five-percent change in attack speed can turn a tank into an assassin. Games, systems, and problems that reward patient analysis of complex, moving parts all pull at the same thread for me.

Get in touch

I’m always happy to talk systems, AI, or anything in between — reach me at contact@yanxitao.com.