Trading technology, verified.

JATS

An event-driven Java platform for developing, backtesting and trading automated strategies. Licensable, and the core product.

Most tools hide execution behind an interface. Six months in you still cannot say why your backtest drifts from live, or why slippage ate your edge. JATS exposes every layer — matching engine, latency model, condition filters, execution algorithms, per-bar state — and proves they agree.

Tick-level backtest ↔ live parity	verified at $0.00 across thousands of trades
Execution algorithms	adaptive implementation shortfall, netting, slot manager, crossing limits
Machine-learning inference	in the decision loop, not beside it
Connectivity	financial messaging protocol, broker abstraction across gateways

A human and an algorithm at the same desk. Click trading, order management and a consolidated portfolio view let a trader work alongside a running strategy. That is the institutional execution-platform pattern — not the retail one, where you deploy and walk away.

Deployed on your infrastructure or in the cloud. We handle the connectivity end to end.

Demo and test installation. Ask by e-mail for a demonstration — or for a test installation on Amazon Web Services that you can work with yourself.

jawata

Compiler-accurate Java analysis and refactoring for AI agents. Open source, sponsored by Quantefakt, because our main product is written in Java — and an agent that edits Java as text silently breaks what it cannot see.

Task: find every call to UserService.save()	Result
grep "save("	47 hits — including orderService.save(), saveButton, comments
jawata	exactly 12 — the real calls

Enforcement. Other tools hope the agent uses them. jawata makes sure: reach for text search on Java and you are blocked, and handed the right call instead.

Grounded memory. Portable files in your repository, anchored to your symbols through the compiler. Ask by symbol, get the lesson written for it — or an honest nothing. Never a pile to sift.

Verification. Every change runs through the same engine the Eclipse development environment uses, is checked against the compiler, and reverses in one call.

For twenty years Java's indictment was that it was slow to write. The AI era zeroes out the cost of writing — the model writes everything, in any language, at the same speed. What is left priced is checking, and there the properties that made Java verbose are the assets. You no longer choose a language for the writer. You choose it for the verifier.

The project has its own site: jawata.org →

What we have built

ETF arbitrage. Fair value computed live for exchange-traded funds and every constituent. Arbitrage surfaced as it opens. Automated order generation and build-and-unwind algorithms across multi-leg correlated baskets — with a click trader for working the illiquid legs by hand. In continuous production since 2015, a nine-figure book, a thousand tickers of live prices.

Execution — we built the matching engine. Almost every backtest in this market prices a fill as bar close plus a slippage constant. Ours matches against historical quotes and ticks, so the answer is whether that order would actually have filled, and at what price.

Patience pays on the way in and costs on the way out. The adaptive limit adds +6.45 percentage points on entry by capturing the spread — and costs −1.51 on exit, because once a stop has fired, price is by definition moving against you and waiting only widens the loss. So: adaptive in, cross out. You can only find that if you own the matching engine.

85 trading days, January–May 2026. Backtest figures about the platform, not a track record: past performance does not indicate future results.

ORB. A systematic intraday equity strategy, implemented for a client on JATS and live on the client's side: a deep-learning ranker feeding the execution stack above. The claim worth making is parity, not return — its selection in JATS is bit-exact against an independent simulator.

Quantitative investment funds. Spreadsheets replaced by a tested calculation core with holdings storage, overlay management and duration steering, delivered as a spreadsheet plug-in so the desk kept its habits. Around 100 funds. Large fund management company.

Co-location futures. The platform extended to a distributed architecture with extended risk management, running in co-location. Large pension fund, North America.

Reinforcement-learning strategies. Long-short and long-only S&P 500 strategies on an in-house library; an intraday strategy for a crypto perpetual contract. Proprietary.

Pairs trading, and VWAP mean-reversion scalping. In slow, balanced markets price is driven by flow benchmarked to the volume-weighted average, so the trade is the rebound to it off a quick deviation that hits resistance — rising volume, an iceberg. Proprietary.

What we do for clients

Maintenance and support. For JATS and for everything we supply. Strategies have to work for years, and you need to know help arrives fast when something unexpected happens.

Strategy development. Arbitrage, implementation-shortfall and market-impact reduction, market making, long-horizon investment. On JATS, or integrated into your own platform.

Java development. Platform, gateway, connectivity, execution-algorithm and risk work — built through a verification layer rather than an unaudited agent loop.

AI and machine-learning consulting. Reinforcement learning for strategy learning and for execution-cost minimisation. Including the honest feasibility answer, if that is what you need.

For fund management companies, asset managers, hedge funds, brokers and proprietary trading firms.

We. Deliver. Always. Trust is the feeling that something sensible still happens when you are not watching.