- Trading infrastructure. Matching engines, order gateways, market data paths, execution and risk in the hot loop.
- Post-trade systems. Securities settlement, payments, deterministic straight-through processing, reconciliation at scale.
- Performance engineering. JVM tuning, GC and allocation analysis, mechanical sympathy, kernel-bypass networking, NUMA and CPU pinning.
- Architecture review. Latency, correctness, and capacity audits of an existing system or a system on the design table.
Rüishi Lab
Rüishi Lab is the practice through which I take on independent consulting engagements in low-latency systems and capital markets infrastructure. The work is narrow on purpose: deterministic execution, zero-GC Java, lock-free concurrency, market data and matching engines, post-trade settlement and payments.
I take engagements where the bottleneck is depth rather than headcount — where one engineer who has actually owned the latency budget can change the answer. If the problem is organizational, or the answer is more people, I'm not the right call.
Audit
Two to four weeks. A structured review of a single subsystem — a matching engine, a market data path, a settlement leg. Output is a written report with prioritized findings and, where appropriate, a working patch.
Embedded engagement
Three to six months alongside an internal team, owning a specific scoped problem: a hot-path rewrite to a sub-microsecond budget, a deterministic STP pipeline, a system taken from prototype to production.
Advisory retainer
Ongoing technical advisory for engineering leadership. Monthly cadence, fixed retainer, written notes. Useful where the constraint is not capacity but a regular second pair of eyes from someone who has seen the failure modes before.
Banks, exchanges, proprietary trading firms, market makers, and the small set of fintech infrastructure companies serving them. I do not take engagements that conflict with my current employment or with NitroJ, and I'm explicit about this before any conversation goes far.
Email rueishi@gmail.com with a paragraph about the problem and the rough timeline. The faster I can tell whether I'm the right person, the less of your time I waste.