Generate, backtest, validate.
Candidate strategies are produced and tested in a continuous loop, with statistical validation gating what gets promoted forward.
Est. 2025 · Brisbane, Australia
Self-improving research and execution engines for futures markets.
Axial Engineering is an Australian quantitative trading firm. We build statistical and quantitative research engines, and the execution infrastructure that runs the strategies they produce.
What we build
Two systems, designed to operate together and to improve themselves over time with minimal manual intervention.
A research engine that generates, backtests, and statistically validates candidate strategies; and an execution engine that runs the strategies that survive validation against live futures markets. Both are self-maintaining: they monitor their own health, recalibrate on new data, and flag degradation without waiting to be asked.
Run validated strategies on futures.
Live execution against CME-listed micro futures, with position management, risk controls, and continuous monitoring.
Recalibrates and reports itself.
Performance, drift, and infrastructure health are measured on a fixed cadence. Degraded strategies are retired automatically.
AI & ML infrastructure
Axial Engineering's research and execution stack runs on modern AI and machine-learning infrastructure. We combine our own quantitative work with industry-leading foundation models and learning systems so a small team can cover a wide research surface and operate at a cadence a statistics-only workflow could not.
Anthropic Claude sits at the centre of the research loop, used to read, reason about, and propose new directions across the quantitative-finance literature. Machine-learning components shape live execution and adapt to current market conditions. Vector-embedding infrastructure helps the system understand similarity and coverage across the research corpus. The internal methods are kept deliberately under the hood; what’s public is the shape of the stack, not the recipe.
Powered by Anthropic Claude.
Frontier large-language models drive how we survey literature, form hypotheses, and prioritise where research effort is spent — running on Anthropic’s production infrastructure.
Adaptive trade gating.
Machine-learning components sit alongside the execution engine and adjust which signals fire and how aggressively they size, in response to live market conditions.
Coverage-aware exploration.
Embedding infrastructure from Voyage AI indexes the research corpus so the system can search by meaning and steer toward less-explored regions of the strategy landscape.
Market focus
We trade CME-listed micro futures. The product universe is narrow on purpose — it keeps costs, slippage, and regime behaviour measurable, which is what the research engine needs to be honest about what is and isn’t working.