Model Platform

Deploy, govern, and monitor ML models on Kubernetes — with built-in EU AI Act compliance.

---

What is Model Platform?

Model Platform is an open-source MLOps platform that lets ML engineers version, deploy, host, and govern machine learning models on Kubernetes with minimal configuration. It bridges the gap between model training (MLflow) and production serving — while generating the compliance documentation that the EU AI Act now requires.

Why?

Deploying ML models to production is still manual and fragmented. Most teams glue together scripts, CI pipelines, and custom tooling — with no audit trail and no governance story.

Meanwhile, the EU AI Act (effective 2025) requires documentation, traceability, risk classification, and human oversight for AI systems. Compliance can't be an afterthought bolted onto spreadsheets.

Model Platform solves both: one platform for deployment and governance.

Key Features

• One-click deployment — Push any MLflow model to Kubernetes with a single action. Auto-provisioned namespace, service, and ingress.
• EU AI Act compliance cards — Auto-generated regulatory documentation per model: risk classification, Article 11 technical docs, traceability records.
• AI-assisted compliance review — Claude analyzes your model card and generates AI Act compliance assessments.
• Per-project namespace isolation & RBAC — Each project gets its own Kubernetes namespace with role-based access control.
• Governance audit export — One-click ZIP export of all compliance artifacts for regulatory review.
• Automatic monitoring dashboards — Grafana dashboards auto-provisioned per deployed model (latency, throughput, errors).
• Full-text search across model metadata — Search across all model cards, descriptions, and metadata from a single search bar.

Screenshots

Projects Overview

Multi-project platform with governance scope visible at a glance — project cards show owner, deployed models, and MLflow status.

EU AI Act Governance

Auto-generated EU AI Act compliance cards with risk classification, Article 11 documentation, and traceability — the unique differentiator.

Model Deployment

Deploy and monitor model in one click

Architecture

C4 Container diagram — shows the major containers and their interactions within the Kubernetes cluster.

graph TB
    User([👤 User])
    CLI([⌨️ CLI — Typer])

    User -->|model-platform.com| Ingress
    CLI -->|REST API| Ingress

    subgraph K8s["☸ Kubernetes Cluster"]

        subgraph NS_Default["default namespace"]
            Ingress[NGINX Ingress]
            Proxy[NGINX Reverse Proxy]
            Ingress --> Proxy
        end

        subgraph NS_MP["model-platform namespace"]
            Frontend["Frontend<br/>HTML/CSS/JS<br/>:80"]
            Backend["Backend<br/>FastAPI<br/>:8000"]
        end

        subgraph NS_Data["Shared Infrastructure"]
            PostgreSQL[("PostgreSQL<br/>Metadata & Users<br/>:5432")]
            MinIO[("MinIO<br/>S3 Artifacts<br/>:9000")]
        end

        subgraph NS_Monitoring["monitoring namespace"]
            Grafana["Grafana<br/>Dashboards<br/>:80"]
            Prometheus["Prometheus<br/>Metrics<br/>:9090"]
        end

        subgraph NS_Project["project namespace (1 per ML project)"]
            MLflow["MLflow<br/>Model Registry<br/>:5000"]
            Model1["Model Service A<br/>:8000"]
            Model2["Model Service B<br/>:8000"]
        end

        Proxy -->|"/"| Frontend
        Proxy -->|"/api/"| Backend
        Proxy -->|"/registry/{project}/"| MLflow
        Proxy -->|"/deploy/{project}/{model}/predict"| Model1
        Proxy -->|"/grafana/"| Grafana

        Backend -->|"SQL"| PostgreSQL
        Backend -->|"K8s API"| NS_Project
        Backend -->|"HTTP"| Grafana
        MLflow -->|"S3"| MinIO
        Model1 -.->|"metrics"| Prometheus
        Model2 -.->|"metrics"| Prometheus
        Prometheus --> Grafana
    end

    Backend -->|"AI Act review"| Claude["Claude AI<br/>Compliance<br/>(external)"]

    style K8s fill:none,stroke:#326CE5,stroke-width:2px
    style NS_Default fill:#E8F5E9,stroke:#009639
    style NS_MP fill:#E3F2FD,stroke:#0E2356
    style NS_Data fill:#FFF3E0,stroke:#E65100
    style NS_Monitoring fill:#FBE9E7,stroke:#F46800
    style NS_Project fill:#E0F7FA,stroke:#00A3BE,stroke-dasharray:5
    style Frontend fill:#00A3BE,color:#fff
    style Backend fill:#0E2356,color:#fff
    style Proxy fill:#009639,color:#fff
    style Ingress fill:#009639,color:#fff
    style PostgreSQL fill:#336791,color:#fff
    style MinIO fill:#C72C48,color:#fff
    style MLflow fill:#0194E2,color:#fff
    style Grafana fill:#F46800,color:#fff
    style Prometheus fill:#E6522C,color:#fff
    style Claude fill:#D97706,color:#fff
    style Model1 fill:#00A3BE,color:#fff
    style Model2 fill:#00A3BE,color:#fff

Loading

Quick Start

See HOWTO.md for full setup instructions (Minikube, Helm, secrets, deployment).

# TL;DR
brew install minikube kubectl helm
minikube start --cpus 2 --memory 7800 --disk-size 50g
make k8s-infra
make create-backend-secret POSTGRES_PWD=... JWT_SECRET=... ADMIN_EMAIL=... ADMIN_PWD=...
make k8s-modelplatform
# → http://model-platform.com

Tech Stack

Component	Technology
Backend API	FastAPI (Python, Clean Architecture)
Frontend	HTML/CSS/JS (served by NGINX)
CLI	Typer (mp command)
Model Registry	MLflow
Database	PostgreSQL
Object Storage	MinIO (S3-compatible)
Orchestration	Kubernetes (Minikube for dev)
Monitoring	Grafana / Prometheus
AI Compliance	Claude (Anthropic)

Documentation

• HOWTO.md — Setup & deployment guide
• CONTRIBUTING.md — Contribution guidelines
• docs/ — Architecture decisions, model card templates, and more

---

Built by OCTO Technology