← Home
On Development · ClassifiedN — 01

Institutional Terminal
for AI-Powered Derivatives.

AI-Powered · Google Gemma 4+FinBERTAGENT
In partnership withFloat Infinity

Nexus is an access-restricted institutional derivatives terminal — a single surface for execution, risk, research, and intelligence. Seven precision- engineered layers route raw exchange feeds from Binance, OKX, and MEXC through battle-tested controls: multi-model VaR, vol-adjusted Kelly sizing, market stress scoring, and a real-time circuit breaker. Every signal, threshold, and allocation is auditable, deterministic, and fortified by live AI synthesis.

Layers7
VenuesBinance · OKX · MEXC
AI StackGemma 4 + FinBERT
Cadence2s Refresh
Active Exchanges3Binance · OKX · MEXC
Alpha Signals11Multi-Timeframe Composite
Risk Modules5VaR · Kelly · Stress · Breaker
AI ModelGemma 4with FinBERT Sentiment
01Low-Latency Multi-Venue Ingestion

Built a multi-venue derivatives terminal ingesting full-depth order book data from Binance, OKX, and MEXC with 100ms–2s signal latency, implementing cross-venue book fusion for unified market state reconstruction.

02Institutional Risk & Execution Framework

Engineered a production-grade risk module combining multi-model VaR, vol/correlation-adjusted Kelly sizing, market stress scoring, and an event-driven circuit breaker for real-time control and dynamic position throttling.

03AI-Powered Intelligence Pipeline

Designed a system-level intelligence pipeline with Gemma 4 narrative synthesis and FinBERT-scored news routing for anomaly detection, regime tagging, and adaptive risk throttling across all active positions.

Execution Terminal
N — 02Layer 01 / 02

Execution Terminal

Unified institutional execution surface — multi-venue candlestick feed with Ichimoku and momentum overlays, full-depth order book, watch zones, and live bias scoring across Binance, OKX, and MEXC.

Risk Module
N — 03Layer 04

Risk Module

Position sizer with vol-adjusted Kelly fraction, multi-model VaR ensemble (historical, parametric, Monte Carlo), market stress index, and an event-driven circuit breaker for real-time exposure control.

Research & Intelligence Brief
N — 04Layer 03

Research & Intelligence Brief

AI-synthesised market brief — Gemma 4 fuses 50+ headlines with funding rate, squeeze risk, macro gate, regime classification, and cumulative volume delta across timeframes.

Alerts & News
N — 05Layer 03

Alerts & News

Sentiment-ranked news feed with Smart Money Index gauge and Fear & Greed read — every headline scored by FinBERT and routed by impact priority.

System Configuration
N — 06Layer 07

System Configuration

Operator console with full pipeline reference — active exchanges, alpha signals, risk modules, fusion engine state, AI model in use, and update cadence across every layer.

L01
Data IngestionInfrastructure

Full-depth order book data ingested from Binance, OKX, and MEXC with 100ms–2s signal latency. Cross-venue book fusion normalises feeds into a unified schema before entering the pipeline.

L02
Signal GenerationAnalytics

Technical and statistical signals on normalised tick data. Multi-timeframe momentum, volume profile, and microstructure features extracted with deterministic replay guarantees.

L03
Intelligence & Bias EngineDecision

Aggregated directional bias from cross-timeframe signal confluence, enhanced by Gemma 4 narrative synthesis and FinBERT sentiment scoring for anomaly detection and real-time alignment.

L04
Risk LayerControl

Institutional risk framework: multi-model VaR ensemble, vol/correlation-adjusted Kelly sizing, market stress index, and an event-driven circuit breaker for adaptive exposure throttling.

L05
Derivatives PricingValuation

Options and perpetual pricing models covering IV surface construction, funding premium, and basis spread relative to spot — updated on each tick with latency-aware validation.

L06
Execution LogicOperations

Order routing, slippage estimation, fill-quality tracking, and venue selection weighted by current liquidity depth and fee tier across all connected exchanges.

L07
Position Sizing & AuditOutput

Kelly-derived sizing with vol-scaled adjustment, hard-capped by the risk layer constraints. Final allocation output is audit-logged and reconciled against real-time exposure tracking.

Part of Louis Bernal's Portfolio

← Back to Portfolio