systems engineer, currently dabbling in inference systems, high perf computing and agentic workloads.
mostly obsessed with the machinery behind intelligence: how a pile of matmuls turns into useful behaviour, how agents turn that behaviour into useful work, and how to make the systems around it fast, reliable, and efficient.
this site is a log of things i'm learning, building, or overthinking. usually the latter.
experience
-
working across agentic workflows, autoresearch, and inference systems on cloud runtimes
-
built backend systems including a risk moderation engine and secure file-sharing system
projects
-
kurama
rust, ratatui, agents
minimal agentic coding harness with a low compute and token footprint
-
sakura
c++, sixel
a high-performance terminal multimedia renderer
-
web server
c, networking
event-driven web server following nginx-style worker architecture
-
vcs
ocaml, vcs
git-like version control system built from scratch
-
load balancer
rust, networking
load balancer built from scratch
-
paimon
python, agents
an agentic coding harness that works across model providers
all projects