Introducing AIVEX: AI Signal Research for Serious Researchers
Why We Built This
The quant research community has a tooling problem. On one side, there are enterprise platforms that are expensive, opaque, and impossible to extend. On the other side, there are scripts — ad-hoc Python notebooks that work until they don't, with no logging, no governance, and no reproducibility guarantees.
AIVEX sits in the gap: a production-grade AI research infrastructure you can inspect, extend, and trust.
What AIVEX Is
AIVEX is a modular, parallel AI signal generation system. It ingests data from multiple independent sources — financial news, OHLCV market data, fundamental metrics — and runs each through a dedicated analysis module. The outputs are then composed, conflict-resolved, and governed before reaching the API surface.
Every step is logged. Every exception generates a structured JSON crash record. Every signal carries a confidence score.
The Five Core Modules
News Signal scrapes and analyzes financial news from 8 sources. An NLP pipeline extracts sentiment, named entities, event types, and urgency signals in real time.
Chart Signal computes 8 technical indicators across multiple timeframes. It detects patterns, trends, and divergences using configurable lookback windows.
Metrics Signal processes fundamental data — earnings, valuation ratios, analyst estimates, and macro indicators — from verified, timestamped sources.
Signal Engine (Brain) is the composition layer. It merges atomic signal outputs using confidence weighting and applies cooldown logic to prevent signal churn. Conflicting signals are resolved via configurable priority rules.
Governor is the risk gate. Before any signal reaches the API, it passes through the Governor, which applies thresholds, kill switches, and human-override policies. No signal is emitted without explicit approval from this layer.
What Makes It Different
Most AI research tools are prediction machines. AIVEX is an evaluation machine: it produces auditable research artifacts, not trading recommendations.
This distinction matters. Every output from AIVEX includes:
- A confidence value in [0, 1]
- A traceable computation path
- A timestamp and source attribution
- A research disclaimer
The Watchdog module ensures the entire system runs 24/7 with automatic restart on failure, crash-loop detection, and heartbeat monitoring.
What's Next
AIVEX is available to qualified research teams by invitation. The full pipeline — ingestion, composition, governance, and the Eye REST API — is running in production. If you are building systematic research workflows and want to evaluate the system, get in touch via the contact form.
For research and informational purposes only. Not financial advice.
For research and informational purposes only. Not financial advice.