MATIH Platform is in active MVP development. Documentation reflects current implementation status.
1. Introduction
Technology Stack
Overview

Technology Stack Overview

The MATIH Platform is built on a carefully selected set of open-source and cloud-native technologies. Every technology choice is guided by three principles: cloud-agnosticism (no proprietary cloud service lock-in), Kubernetes-native deployment, and production readiness. This section documents every technology in the platform, organized by layer.

Stack at a Glance

LayerTechnologies
BackendJava 21, Spring Boot 3.2, Python 3.11, FastAPI, Node.js 20, Express
FrontendReact 18, TypeScript, Vite, TailwindCSS, Zustand
Data InfrastructurePostgreSQL 16, Redis 7, Kafka (Strimzi), Elasticsearch 8.11, MinIO
ML and AIMLflow, Ray, vLLM, LangGraph, Qdrant, Feast, Triton
Orchestration and IaCKubernetes 1.29+, Helm 3.14+, Terraform, cert-manager, ArgoCD
Compute EnginesTrino, ClickHouse, StarRocks, Apache Spark, Apache Flink

Selection Criteria

Technologies are selected based on the following criteria:

CriterionRequirement
Open-sourcePrefer Apache 2.0, MIT, or similarly permissive licenses
Kubernetes-nativeMust run on Kubernetes with Helm chart or Operator support
Cloud-agnosticNo dependency on any single cloud provider
Production-provenUsed in production by multiple large organizations
Active communityRegular releases, responsive maintainers, active contributor base
Multi-tenancy compatibleSupports tenant isolation at the application or infrastructure level

Technology Versions

Backend

TechnologyVersionPurpose
Java21 (LTS)Control Plane and Data Plane Java services
Spring Boot3.2Application framework for all Java services
Python3.11AI, ML, and data quality services
FastAPILatestAsync web framework for Python services
Node.js20 (LTS)Render service
ExpressLatestWeb framework for Node.js services

Data

TechnologyVersionPurpose
PostgreSQL16Primary relational database
Redis7Caching, sessions, pub/sub
Apache KafkaStrimzi 0.38+Event streaming
Elasticsearch8.11Full-text search, audit logs
MinIOLatestS3-compatible object storage

Compute

TechnologyVersionPurpose
TrinoLatestFederated SQL query engine
ClickHouseLatestOLAP analytical engine
Apache Spark3.5Batch data processing
Apache Flink1.18+Stream processing
Ray2.9+Distributed ML compute

Infrastructure

TechnologyVersionPurpose
Kubernetes1.29+Container orchestration
Helm3.14+Package management
Terraform1.6+Infrastructure as code
cert-managerLatestTLS certificate management
NGINX IngressLatestPer-tenant ingress controller
External Secrets OperatorLatestSecret synchronization
PrometheusLatestMetrics collection
GrafanaLatestMonitoring dashboards

Polyglot Architecture

The platform uses three programming languages, each selected for its strengths:

LanguageServicesRationale
Java 21Control Plane (all 10), Query Engine, Catalog, BI, Pipeline, Semantic LayerStrong typing, mature ecosystem, excellent Spring Boot support, Hibernate multi-tenancy
Python 3.11AI Service, ML Service, Data Quality, Governance, Ontology, Ops AgentML/AI ecosystem (LangChain, Ray, scikit-learn), FastAPI async performance
Node.js 20Render ServiceChart rendering libraries (D3.js, Chart.js), PDF generation

Commons Libraries

Shared functionality is extracted into four commons libraries:

LibraryLanguageKey Modules
commons-javaJavaSecurity, multi-tenancy, persistence, caching, event streaming, observability
commons-pythonPythonAuthentication middleware, tenant context, structured logging
commons-typescriptTypeScriptAPI client utilities, authentication hooks, shared UI components
commons-aiPythonLLM abstractions, prompt management, RAG utilities, agent framework

Next Steps

Explore each technology layer in detail:

Technology StackBackend Stack