Systems
Applied work and system-level thinking across trading infrastructure, AI data flows and adaptive control.
Trading systems
Trading systems provide a useful test case for AI because the constraints are explicit: sequencing, state, latency, auditability and bounded risk. Statistical models can identify patterns, but execution and validation remain deterministic.
FIX and structured message flows
Message streams as sequence data, with schema, state and lifecycle semantics that should not be treated as ordinary text.
Model-assisted operations
Anomaly detection, support summarisation, reject classification and operational monitoring.
Adaptive systems
Blood glucose control is a real-world control problem with noisy input, delayed interventions and safety limits. It is a useful counterpoint to trading systems because the constraints are physiological rather than regulatory.
Insulin action and GIR curves
Using insulin action profiles to reason about dose timing, basal coverage and transitions between dosing regimes.
Prediction versus control
Short-term prediction may be useful, but safety depends on bounded adjustment, confidence and explicit limits.