Infrastructure Deep Dive

Institutional Execution
Architecture

12+ months of engineering, LP negotiations and infrastructure build — this is what separates a strategy that runs in production from one that runs in a backtest.

From Signal to Fill

Every trade travels through a purpose-built institutional execution chain before reaching liquidity

AI Algorithms
ML/RL signal generation
Metaquotes
MetaTrader 5 execution engine
FIX 4.4 Bridge
MT5 → institutional protocol
PrimeXM
Aggregation & smart routing
FIX API
Industry-standard protocol
Routing ~20 LPs
Smart fragmentation& best execution

Layer 01 — MT5 + ML/RL Engine

MT5 serves as the execution engine for our proprietary AI algorithms. Custom MQL5 code interfaces with institutional infrastructure — far beyond standard retail MT5 usage.

  • Custom MQL5 expert advisors — proprietary logic
  • Adaptive ML/RL signal generation
  • Real-time risk parameters at EA level
  • Portfolio: XAUUSD · NAS100 · US30 · US500 · GER40

Layer 02 — FIX Bridge

The critical translation layer between MetaTrader's proprietary protocol and the institutional FIX ecosystem, where retail solution ends and institutional execution begins.

  • Bridges MT5 order flow to FIX 4.4 protocol
  • Sub-millisecond latency at bridge layer
  • Order type translation — market, limit, IOC
  • Configured with Metaquotes

Layer 03 — PrimeXM

The institutional aggregation layer between our FIX Bridge and the LP network. Handles smart order routing, LP stream aggregation and real-time best execution logic.

  • Real-time aggregation of ~20 LP price streams
  • Smart order routing best bid/ask selection
  • Fill quality monitoring + LP performance scoring
  • Enterprise-grade uptime institutional SLA

Layer 04 FIX API to ~20 LPs

Outbound FIX connections to each liquidity provider — the same protocol used by tier-1 prime brokers globally. Dedicated sessions per LP with isolated failover.

  • Dedicated FIX session per LP — isolated failover
  • Heartbeat monitoring dead session detection
  • Fill confirmation + latency profiling per LP
  • No single LP >15% of flow at $50M AUM

The LP Network — The Hardest Part

Not all LPs are equal. Finding providers comfortable with our execution profile took months of live testing, spread analysis and negotiation. 50+ tested, ~20 retained.

  • Selected for tolerance to high-frequency directional flow
  • Spread tightness profiled across sessions + instruments
  • No progressive widening after sustained flow
  • Capacity design: diversified to prevent saturation

Risk & Monitoring Layer

Real-time monitoring across the full stack — from EA-level position limits to LP-level fill quality. Live visibility into every account and execution metric.

  • Per-account drawdown limits enforced in real time
  • LP fill rate + slippage tracking — automated alerts
  • Live dashboard
  • AMD Epyc + Supermicro infrastructure on Vultr

12 Months to Production

What it actually took to build an institutional execution stack from scratch

Month 12

Metaquotes relationship & MT5 customisation

Establishing a working relationship with Metaquotes to access institutional-grade MT5 server infrastructure and FIX Bridge documentation. Custom EA development and adaptation for low-latency institutional execution.

Month 2–3

PrimeXM onboarding & FIX configuration

PrimeXM account setup, FIX Bridge integration and initial LP stream configuration. Extensive testing of order routing logic, fill quality at different lot sizes and latency profiling across instruments.

Month 3-8

LP selection — the hardest part

Identifying liquidity providers willing to sustain our execution profile without progressive spread widening or flow toxicity rejection. Tested 50+ LPs in live conditions. Retained ~20 that met fill quality and tolerance criteria.

Month 8–12

Algorithm adaptation & go-live

Adapting ML/RL algorithms to the production execution environment accounting for real-world latency, partial fills and LP-specific execution characteristics. First live client account: April 2024.

What makes this defensible

This infrastructure cannot be replicated in weeks. The combination of Metaquotes relationships, PrimeXM configuration expertise and a curated LP network tolerant of our execution profile represents a full year of work and institutional relationships that are not publicly available.

Any partner integrating with this stack inherits a production-ready institutional execution layer — not a demo environment.

Result after 25 months live

$54k→$714k
Account from dashboard - single account
$7.6M
Total AUM - spread on 5 accounts
2.88x
Profit factor - 25 months
8.17%
Max drawdown

Designed to Scale

Every architectural decision was made with capacity in mind from $54k to $500M+

$1.2M
Starting AUM
Starting capital deployed across first live production accounts. April 2024.
Live - achieved
$7.6M
Current AUM
Profit across 5 accounts. Different risk profiles & instruments over 25 months.
Live - today
$50M
Config ceiling
Current LP network supports up to $50M without execution degradation.
Next milestone
$500M+
Tier-1 PB phase
Beyond $50M: tier-1 prime brokers (Citadel Sec...) to maintain fill quality.
Roadmap

Why $50M is the empirical ceiling — not theoretical

The $50M figure comes from LP depth analysis across our ~20 providers at current execution sizes. Beyond that threshold, queue positioning deteriorates and spread widening becomes measurable on XAUUSD and indices during peak sessions. This is a live observation from production data — not a model assumption. The tier-1 PB onboarding roadmap is specifically designed to add depth at that transition point.

What You Can't Buy Off the Shelf

Institutional execution infrastructure is not a commodity each element required time, relationships and iteration

Metaquotes relationship

Direct working relationship for FIX Bridge access and institutional MT5 server configuration — not available through standard channels.

PrimeXM configuration expertise

Months of live tuning of routing logic, LP stream weighting and fill quality parameters institutional knowledge embedded in the config.

Curated LP network

~20 LPs selected from 30+ tested in live conditions for tolerance to high-frequency directional flow. These relationships are not publicly replicable.

25 months of live adaptation

Algorithms adapted to this specific execution environment latency, partial fills, LP-specific behaviour. That adaptation is not portable to generic infrastructure.

Proven capacity design

$54k $714k single account, $7.6M total AUM — execution quality validated at increasing scale in live conditions, not modelled.

Full vertical integration

Signal generation, execution, risk monitoring and reporting in one integrated stack — no dependency on third-party black boxes at any critical layer.

Next Step — Dark Factory

The next evolution: a fully autonomous institutional trading system

Autonomous Operation with Archon OS

The next evolution of this infrastructure is full autonomous operation a Dark Factory model where the entire stack runs without human intervention (almost). Signal generation, execution, risk management and reporting operate as a self-managing agentic system 24/7.

We are integrating Archon OS — an open-source agentic AI framework (13.7k+ GitHub stars) — as the knowledge and task management backbone. Archon orchestrates AI agents across the full trading pipeline: strategy monitoring, anomaly detection, LP performance alerts and autonomous parameter adaptation.

The result: a fully autonomous institutional trading system that self-monitors, self-optimises and self-reports — with human oversight reserved for strategic decisions only.

Partner with the Infrastructure

We are building Nexus Fund with the right institutional partner. If you understand what this infrastructure represents let's talk.