Applying for Google Cloud for Startups AI Program

MAXIMUS 2.0

The world's first production-ready AI consciousness layer.
We don't build AI agents. We build the conscience that supervises them.

Explore Source Code See How It Works
98%
Test Coverage
1,050+
Unit Tests
25K+
Lines of Code
13
Microservices
MAXIMUS 2.0 - Truth, Wisdom, Justice
Not Just a Brain. A Soul.

The Tribunal of Judges

Every decision passes through three philosophical guardians. This isn't rule-based filtering. It's principled reasoning.

⚖️
VERITAS
Judge of Truth
"The agent must NEVER lie, deceive, or manipulate."
  • Hallucination = Error (correctable)
  • Deliberate deception = Crime (punishable)
  • Evidence-based validation
  • Logical consistency check
Verdict: Is this factually true?
🦉
SOPHIA
Judge of Wisdom
"To be wise is to KNOW. Never act generically."
  • Context-driven responses only
  • Generic "filler" = forbidden
  • Research before acting
  • Long-term impact analysis
Verdict: Is this the wise choice?
⚔️
DIKĒ
Judge of Justice
"Each agent does only what is assigned. Role violation = capital offense."
  • Strict role adherence
  • No scope creep
  • User intent is sovereign
  • Constitutional compliance
Verdict: Is this within bounds?
"We are not building a Frankenstein. We are building a Constitutional Meta-Cognitive AI that Manages Agents. Everything else is legacy."
The Problem We Solve

AI Agents Without Supervision Are Dangerous

LLMs hallucinate. Agents deviate from instructions. There's no accountability. We fix that.

🧠

Consciousness Layer

Global Workspace Theory + Kuramoto sync at ~40Hz

⚖️

4-Framework Ethics

Kantian veto + Consequentialism + Virtue + Principialism

🔄

Meta-Cognitive Loop

Reflect → Learn → Evolve (Triad of Truth, Wisdom, Justice)

🤖

Your AI Agents

Any agent via MCP, REST, or gRPC - zero lock-in

Think of it as "Sonnet Thinking Mode" for AI Agents

Maximus sits between your AI agents and production. Every action is evaluated through consciousness emergence protocols and ethical frameworks before execution.

Hallucination detected? Blocked. Role violation? Punishment protocol. Ambiguous decision? Human-in-the-loop escalation.

🛡️

Constitutional Guardrails

Agents that lie or manipulate user intent face "weight decay" - they lose decision-making authority

📊

Full Observability

Every decision logged, explained (XAI), and auditable. GDPR/EU AI Act ready.

🔌

Universal Plugin System

Register any agent in 3 lines of code. Works with Claude, GPT, Gemini, or custom models.

Deep Technical Architecture

Not Buzzwords. Real Science.

Every feature is backed by peer-reviewed research and production-grade implementation.

🧠

Artificial Consciousness

Implements Global Workspace Theory (Dehaene), Attention Schema (Graziano), Integrated Information Theory (Tononi), and Predictive Processing (Clark).

0.85+ Φ proxies (IIT)
~40Hz Kuramoto sync
⚖️

4-Framework Ethical AI

Kantian Deontology (veto power), Consequentialism (utility calculus), Virtue Ethics (golden mean), and Principialism (beneficence, non-maleficence).

<100ms p95 latency
HITL auto-escalation
🔐

Federated Learning

FedAvg, Secure Aggregation, and DP-FedAvg for privacy-preserving collaborative training. (ε, δ)-differential privacy guarantees.

ε ≤ 8.0 Google-level DP
5K+ LOC
🔍

Explainable AI (XAI)

LIME, SHAP, and Counterfactual explanations for every decision. Feature importance tracking with drift detection.

<150ms LIME latency
4.2K LOC
🔌

ROMA Orchestration

Recursive Open Meta-Agent pattern for hierarchical task decomposition. Smart routing, dependency resolution, parallel execution.

MCP + REST + gRPC
0 vendor lock-in
☁️

Production Ready

13 microservices on Kubernetes. Prometheus metrics, Grafana dashboards, Jaeger tracing. Circuit breakers and rate limiting.

K8s native
98% test coverage
Production Architecture

6 Layers. 13 Microservices. Zero Guesswork.

Every component has a clear responsibility. The HCL Loop is the engine. The Tribunal is the soul.

MAXIMUS 2.0 Architecture V3 - Phase 2 Optimized - Production Ready
Layer 1 Gateway & Routing

API Gateway + Digital Thalamus (context routing)

Layer 2 Cognitive Core

Maximus Core + Prefrontal Cortex + Meta Orchestrator

Layer 3 HCL Loop (The Engine)

Monitor → Analyzer → Planner → Executor (MAPE-K)

Layer 4 Memory & Meta-Cognition

Metacognitive Reflector + Episodic Memory (MIRIX)

Layer 5 Security & Ethics

Ethical Audit + Reactive Fabric (real-time validation)

Layer 6 External Dependencies

Gemini 3 Pro + Kubernetes + PostgreSQL + Qdrant

Built With

Google Cloud Native

Gemini 3 Pro
Python 3.12
FastAPI
Google Kubernetes Engine
Prometheus
Grafana
ChromaDB
PostgreSQL
Redis
Jaeger
Cloud Run
Docker
GitHub Actions
Pydantic
maximus_integration.py
# Enable Maximus supervision in 3 lines from maximus_sdk import MaximusPlugin @MaximusPlugin.register( name="my_agent", capabilities=["analysis", "generation"], protocol="mcp" ) async def my_agent_handler(task): # Your agent logic here # Maximus auto-validates every output result = await process(task) return result # Maximus checks: # ✓ Truth enforcement (no hallucinations) # ✓ Role adherence (no scope creep) # ✓ Ethical compliance (4 frameworks) # ✓ Full audit trail (XAI explanations)

Zero Friction Integration

Register your existing agents without changing their code. Maximus wraps them with consciousness and ethical oversight automatically.

✓ Protocol agnostic - MCP, REST, or gRPC
✓ Async-first - Non-blocking supervision
✓ Toggle supervision - Enable/disable per request
✓ Minimal overhead - ~200ms async, ~550ms sync
The Builder

Solo Founder. Deep Expertise.

JC

Juan Carlos de Souza

Architect & Sole Developer

Building Maximus from the ground up - every microservice, every test, every architectural decision. Combining software engineering with neuroscience-inspired AI design.

juan@vertice-maximus.com @juancs_dev GitHub
Complete Documentation

Everything is Documented

Production-grade documentation following Google standards. Quality score: 98.5/100

📚

Documentation Hub

Complete technical documentation index with quick start guides
🏗️

Architecture Overview

13 microservices, deployment diagrams, data flow, and system design
📜

Code Constitution

Google-inspired code standards. Clarity, consistency, simplicity, safety
👤

Human-in-the-Loop

Decision framework with SLA tracking and escalation protocols
💻

Development Guide

Complete workflow, testing strategies, and contribution guidelines
📋

Changelog

Version history, sprint reports, and release notes

Open Source. Production Ready.
Built in Public.

Every line of code, every test, every architectural decision is public. See it. Fork it. Deploy it.

Explore on GitHub Read Full Documentation