E.T.S. is a production-grade RL environment that simulates the compliance enforcement pipeline of a HFT. It challenges and agent to detect and classify financial anomalies in microsecond scale trades.
Since I always had an intrest in fintech, we thought of making a project in this domain
This is not a simulation that pretends to be fast. It's a genuine lock-free, cache-friendly data structure running inside your Python process.
The core engine is written entirely in C++20 and exposed to Python via nanobind with zero-copy memory semantics. The RL loop, training harness, and FastAPI server are all Python, but the computation that matters happens at C++ speed.
Making a project this complex with soo many working parts was the biggest challenge we faced.
The reconciliation engine hit 10 million+ orders per second throughput consistently.
We learnt how to connect the python environment to the cpp based recon engine.
the only thing holding ETS back is hardware, soo setting up appropriate hardware to see the full potential of ETS.
Log in or sign up for Devpost to join the conversation.