SuperIntel (private research system)
CorrelateX's private trading-research platform for pairs trading. HedgeVision is its open-source teaser.
Media & Demos
Asset Correlation Heatmap
Price & Spread Analytics
Role
Developer
Period
2025
Category
trading
Overview
SuperIntel is CorrelateX's private research system. It combines cointegration and GARCH analysis with a model ensemble, a TimescaleDB feature store and a ports-and-adapters design. The open-source project HedgeVision (github.com/ayush108108/hedgevision) is a fragment extracted from this stream; SuperIntel has run with a small amount of the founder's own capital and is being hardened before more is committed. It manages no one else's money, is not offered to investors, and gives no investment advice.
Key Highlights
- Supabase shim: a thin proxy that routes SDK calls to local TimescaleDB (read latency under 10 ms, self-reported)
- Ports and adapters: Postgres, in-memory and JSONL adapters behind the same interfaces
- Tiered predictions: statistical models, then an ML ensemble, then an LLM review step
- Circuit breakers for LLM providers with fallback order
- Prometheus metrics and health checks
- Private repository: a preview is available on request
Tech Stack
Links
Summary
On-page overviewThis is a concise summary of the challenges, solution, and outcome for this project. Use the Case Study button above for the full deep dive.
The Problem
The 'Latency-Reliability Paradox'. Early reliance on cloud SDKs introduced unacceptable external latency (~150ms). Furthermore, trading requires speed, but ML training/inference (AutoGluon) is CPU-bound and blocking. I needed a system that was sub-10ms fast but smart enough to debate macro narratives without hanging the event loop.
The Solution
1. A thin proxy that mimics the cloud SDK's call syntax and routes queries to local TimescaleDB hypertables, so ingest scripts did not need rewriting. 2. Ports and adapters with factory-based dependency injection, so the same code runs against a Postgres adapter, an in-memory adapter (unit tests) and a JSONL adapter (audit). 3. Tiered predictions (statistical models, ML ensemble, LLM review) and circuit breakers that fall back across LLM providers when one is unavailable.
The Outcome
A research platform where read latency dropped from roughly 150 ms (cloud SDK) to under 10 ms locally, self-reported. No uptime figure is claimed. SuperIntel has run with a small amount of the founder's own capital and is being hardened before more is committed; it manages no one else's money. It is a personal research build, not a multi-user product.
Team & Role
Solo build. CorrelateX research, authored by its founder.
What I Learned
This project deepened my understanding of Python and FastAPI (Async) and reinforced best practices in system design and scalability. I gained valuable insights into building maintainable software and measuring performance.