Systems · AI · Risk
Kernel-level systems engineer turned AI & risk-control builder. MSU system-programming grad — now shipping multi-agent LLM systems and risk-aware quant tooling that grades its own decisions.
"Everything real turns fake, everything right is wrong."
based in Moscow ↔ CN
Authored a patch merged into the mainline tree (Linux 6.9) that fixes an
infinite loop in the selftests/harness FIXTURE_TEARDOWN path —
torvalds/linux · 72d7cb5.
Selected work
From SQLite-from-scratch to autonomous LLM traders. Full list on GitHub.
A fully autonomous, multi-agent LLM system that runs a bull-vs-bear debate over a real HK + US stock portfolio every morning, enforces hard risk gates, and grades its own calls on a daily scorecard — honestly admitting where it loses to buy-and-hold. Ships a live public dashboard.
Writing a SQLite clone from scratch in C++ — B-tree pager, cursor, and REPL. The most-starred thing I've shared.
A course guide for the CS track at MSU-BIT / MSU Department of System Programming — the resource I wish I'd had as a freshman.
Trajectory
Systems & compilers foundation → product-minded AI and risk work.