Equinox
Infrastructure prototype that connects to fragmented prediction market venues (Polymarket and Kalshi), normalizes them into a canonical internal model, and detects equivalent markets via a three-stage pipeline: LLM category alignment, semantic similarity (MiniLM-L6-v2 + FAISS), and GPT-4o-mini verification. A venue-agnostic routing engine evaluates order fills across matched venues — comparing price, depth, slippage, and fees — then produces explained routing decisions. Built as a systems design exercise in cross-venue normalization under real-world ambiguity.
PythonFAISSsentence-transformersOpenAIPydanticTyper

equinoxpre-recorded
$ Select a command below to run a demo