Multi-Container ML Application using Docker Compose

Container Orchestration & Multi-Service MLOps Deployment

Implemented a multi-container machine learning application where a Flask-based UI service communicates with a Dockerized ML inference service using Docker Compose and service networking. This project demonstrates end-to-end MLOps deployment with container orchestration.

Project Summary

Comprehensive Project Overview

Project Category

Container Orchestration & Multi-Service Application Deployment (Docker Compose)

Industry/Domain

Platform Engineering / Cloud Infrastructure

Domain Focus

MLOps - Model Serving & Multi-Container Application Systems

Key Technologies & Concepts

Core Technologies Used

Docker Compose Keywords

Docker Compose (docker-compose.yml) Multi-Container Application Architecture Service-Oriented Container Design ML Inference Service Container UI / API Gateway Container Inter-Container Networking Docker Bridge Network Service Name-Based Discovery REST API Service Communication End-to-End Model Invocation Stateless Service Containers Port Mapping & Exposure Environment Variable Configuration depends_on (Service Startup Ordering)

Problem & Objective

What problem did this project solve?

Problems Solved

Running the ML inference service and client/UI separately caused manual coordination
Networking complexity and inconsistent startup made end-to-end invocation unreliable
Lack of standardized deployment process for multi-service ML applications

Primary Objectives

Orchestrate multiple containerized services using Docker Compose
Enable reliable service-to-service communication between ML and UI containers
Facilitate end-to-end invocation of machine learning models
Create unified application setup for development and deployment

Solution & Architecture

Architectural Overview

Solution Overview

The solution uses Docker Compose to orchestrate two containerized services: a machine learning inference service and a UI/API gateway service. The architecture enables seamless communication between containers through Docker's internal networking.

Docker Compose provides an internal network where each service is reachable by its service name instead of hard-coded IP addresses, allowing reliable container-to-container communication even when containers are restarted or recreated.

This approach decouples the ML inference logic from the UI presentation layer, enabling independent scaling, maintenance, and deployment of each component while maintaining reliable communication.

Docker Compose Multi-Container Architecture Diagram

1

Source Code & Dockerfiles

2

Docker Images

3

Docker Compose Orchestration

4

Service Containers

5

End-to-End Model Invocation

Key Components

Machine Learning Inference Service: Flask + Gunicorn container with trained model
UI / API Gateway Service: Flask container for client interface
Dockerfiles: Separate blueprints for each service
Docker Compose: Orchestration through docker-compose.yml
Docker Engine: Runtime environment
Inter-Container Network: Docker bridge network for service communication
Docker Hub: Container registry for image distribution

Scalability & Reliability

Services designed as stateless containers enabling horizontal scaling under orchestrators
Docker Compose ensures reliable service startup and networking
Consistent end-to-end invocation during development and testing
Lightweight base images and stateless service design optimize container startup

AI/ML & DevOps Implementation

MLOps Focus and Implementation Details

AI/ML Focus

AI/ML-focused project emphasizing multi-container orchestration of machine learning inference service and client application using Docker Compose. The focus is on service integration and deployment readiness rather than model training.

Models & Pipeline

Deployed pre-trained machine learning inference service as standalone container
Implemented automated service-to-service invocation between UI/API gateway and ML service
Breast cancer prediction model using XGBoost algorithm

Containerization & Orchestration Tools

Docker Engine / Docker Desktop Docker CLI Docker Compose (Multi-container orchestration) Docker Hub (Container Registry) Flask (API & UI Service) Gunicorn (Production WSGI Server) Postman (API Testing & Validation)

Skills & Technologies Used

Technical Proficiency Demonstrated

Primary Skills

Docker & Containerization - Intermediate: Building, managing, and optimizing Docker containers
Docker Compose (Multi-Container Orchestration) - Intermediate: Orchestrating multiple services with dependencies
Machine Learning Model Serving - Intermediate: Deploying and serving ML models in production
Service-to-Service Networking - Intermediate: Configuring container networking and communication

Secondary Tools / Frameworks

Flask (API & UI Service Framework)
Gunicorn (Production WSGI Server)
Postman (API Testing & Validation)
Docker Compose CLI
Linux Command Line

Programming Languages

Python: Primary language for ML service and UI/API gateway
YAML: Docker Compose configuration and Dockerfile instructions

Cloud & DevOps Tools

Docker Engine Docker CLI Docker Compose (docker-compose.yml) Docker Hub (Container Registry)

Docker Compose Components

Service Architecture Details

ML Model Inference Service

Service Name ml-model

Port 9696

Base Image python:3.10-slim

Framework Flask + Gunicorn

ML Model XGBoost (Breast Cancer)

API Endpoint /predict (POST)

UI / API Gateway Service

Service Name ui-app

Port 5000

Base Image python:3.10-slim

Framework Flask

Communication HTTP requests to ml-model

API Endpoint /predict (Proxy)

Service Discovery: UI container calls ML service using http://ml-model:9696/predict - leveraging Docker's internal DNS resolution

Architecture to Dockerfile Mapping

Architecture Component	Dockerfile Mapping	Purpose
Base Runtime	FROM python:3.10-slim	Lightweight Python runtime for ML inference
Working Directory	WORKDIR /app	Sets isolated execution directory inside container
Copy Application & Model	COPY . /app	Copies inference code, model file (.pkl), and configs
Dependency Installation	RUN pip install -r requirements.txt	Installs ML and API dependencies
Network Exposure	EXPOSE 9696	Exposes inference API port
Application Startup	CMD ["python", "predict.py"]	Launches Flask-based ML inference service

Docker Compose Commands

Essential Commands for Orchestration

Command	Purpose
`docker-compose up --build`	Builds (if necessary) and starts all services defined in docker-compose.yml
`docker-compose ps`	Lists status of containers managed by Docker Compose
`docker-compose logs`	Displays combined logs of all services for debugging
`docker-compose down`	Stops and removes all containers and their networks
`docker-compose build`	Builds or rebuilds services without starting them
`docker-compose up <service>`	Starts or restarts a specific service
`docker-compose run`	Starts a one-off task without launching entire stack
`docker-compose exec`	Runs a command on a running container
`docker-compose restart`	Restarts all services defined in docker-compose.yml
`docker-compose down -v`	Stops and removes containers, networks, and volumes

Challenges & Outcomes

Technical Challenges and Resolutions

Key Technical Challenges

Coordinating startup and communication between multiple containerized services
Configuring reliable inter-container networking for end-to-end ML inference
Managing service dependencies and startup ordering
Ensuring consistent API communication between containers

Resolution Strategies

Used Docker Compose to orchestrate services with defined dependencies and shared network
Leveraged service-name-based communication for reliable container-to-container API invocation
Implemented depends_on for proper service startup sequencing
Used Docker's internal DNS resolution for service discovery

Note: Terminal screenshots intentionally include intermediate build errors and corrections to demonstrate real-world debugging, problem-solving, and iterative development during containerization.

Assets & References

Code, diagrams, study material

GitHub Repository

Source code repository containing the multi-container ML application with Docker Compose configuration.

Access Repository

Study Material Resources

Click the button below to open the study materials

Request Study Material