Kamui MM Lab

Inspiration

Market making is a future direction for Kamui. The question is how to explore two-sided quoting when the starting balance sheet is small, while making the limits visible instead of promising returns.

What it does

Kamui MM Lab is a local, entirely synthetic research simulator. A visitor sets a budget, inventory cap, and volatility stop, then replays range, trend, or shock paths. The screen shows bid/ask logic, a learned signal, simulated fills, capital use, marked equity, and quote pauses. A fixed-quote reference runs on the same generated path. No exchange connection, deposits, live orders, or investor data are involved.

How it was built

The browser app uses native JavaScript modules, HTML, and CSS, served by a dependency-free Node.js server. A small regularized linear regression learns from generated order-book imbalance, momentum, and volatility. Its forecast shifts the quote center; deterministic inventory and volatility gates can override it. A cash-and-inventory ledger marks each generated tick. We used the Devpost Learn Skill Pack for the public scope, product requirements, and technical specification in the repository.

Challenges and what we learned

The main challenge was making a low-capital comparison inspectable: both strategies use the same synthetic path and starting capital, while the adaptive strategy must respect cash and position limits. In the generated shock, the volatility gate pauses quoting rather than presenting model output as a reason to keep trading. The exercise shows how model signals and hard risk rules can coexist in a prototype; it does not establish real-market profitability.

Next steps

Test with realistic queue position, venue fees, latency, adverse selection, and independent risk review before considering any live system. This project is a demonstration of a research direction, not an investment product or regulatory approval.

Code and planning documents: https://github.com/SCMLpt/kamui-mm-lab