BMLL, Simudyne launch AI market simulator

BMLL and Simudyne have launched an AI market simulator that replays tick-level historical order and trade data to provide a testing environment for trading strategies, risk models and market-structure research. The platform is designed for training and testing AI models and trading systems against realistic market dynamics.

The product pairs BMLL’s normalized tick-level dataset, which records orders, trades and limit order book events across venues, with Simudyne’s cloud-capable agent-based simulation engine that models the behavior of multiple market participants and infrastructure. Users can replay past market periods at full fidelity or run controlled experiments that change participant behavior, latency conditions or execution rules.

BMLL supplies cleaned and standardized feeds intended to remove gaps and align event timestamps across venues. Simudyne provides the simulation toolkit and compute layer that hosts agent-based models and AI agents, allowing researchers and trading teams to run many scenarios in parallel and to instrument the environment for detailed performance analysis.

The simulator is aimed at quantitative researchers, buy-side and sell-side trading desks, model validation teams and market operators who need to evaluate algorithmic strategies, measure model risk and test market-impact assumptions without using live capital. It supports backtesting of AI-driven strategies under microstructure features such as order book dynamics, hidden liquidity and cross-venue interactions.

Technical features include seeding simulations with specific historical intervals, injecting synthetic orders or shocks, and capturing full event logs for post-run analysis. The environment supports training reinforcement-learning agents and other machine-learning models by providing realistic state transitions and reward signals derived from historical fills and market responses. Simulation runs can be scaled in the cloud to accelerate experimentation and run trade-off studies across multiple parameter settings.

No customer names were disclosed in initial descriptions of the product. The companies described the simulator as a tool to expose models to realistic microstructure behavior and to rare but consequential events that may be underrepresented in aggregated datasets, and to help compliance and model-validation teams generate reproducible test cases for regulatory and internal review.

BMLL began as a provider of tick-level market data focused on normalizing and enriching event-level feeds. Simudyne develops agent-based and synthetic market simulation software used by financial institutions and regulators to study market resilience and strategy interactions. The collaboration combines BMLL’s data hygiene and market coverage with Simudyne’s simulation and modelling capabilities to provide an integrated environment for market testing.