How to Deploy ClearML as a Self-Hosted MLOps Platform on Ubuntu with Docker

·2 views

ClearML is an open-source MLOps platform offering experiment tracking, pipelines, hyperparameter optimisation, and model serving as a self-hosted alternative to AWS SageMaker. A technical guide published on DEV Community walks through deploying the ClearML server on Ubuntu using Docker Compose, with Traefik acting as a reverse proxy across three subdomains for the web, API, and file servers. The setup requires an Ubuntu host with Docker installed, DNS records pointing to the relevant subdomains, and optionally the NVIDIA Container Toolkit for GPU workloads. Once deployed, the stack supports the full machine learning lifecycle, including agent registration, sample experiments, pipeline builds, hyperparameter optimisation sweeps, and model serving. The guide is aimed at teams seeking cost-effective infrastructure control without relying on managed cloud ML services.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Step-by-Step Guide: Dockerizing a Flask API on Ubuntu for Production

A technical guide published on DEV Community walks developers through containerizing a Flask REST API on Ubuntu using Docker and Gunicorn. The tutorial begins with installing Docker natively on Ubuntu 22.04 or 24.04, then clones a real-world Flask project from GitHub as a practical starting point. A key focus is writing an optimized Dockerfile that copies requirements.txt before application code, preserving Docker's layer cache and speeding up rebuilds. The guide replaces Flask's built-in development server with Gunicorn to handle concurrent requests safely in production environments. It concludes with steps to version and publish the final Docker image to Docker Hub, covering environment variable configuration for security.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Strix Open-Source Tool Uses AI Agents to Pentest Your App Before Deployment

Strix is a free, open-source security tool that deploys autonomous AI agents to actively attack a running instance of your application, mimicking real-world penetration testing. Unlike static analysis or dependency scanners, Strix spins up the app in a Docker sandbox and attempts to exploit vulnerabilities rather than just flagging suspicious code patterns. It uses a multi-agent architecture where specialized agents work in parallel, covering issues such as SQL injection, access control flaws, XSS, business logic bugs, and infrastructure misconfigurations. Each reported vulnerability includes a working proof-of-concept exploit, reducing false positives that plague traditional SAST tools. Strix supports major LLM providers including OpenAI, Anthropic, and Google, and can target local directories, GitHub repositories, or remote URLs with a single command.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Layered Architecture Explained: How PHP Apps Are Structured Without Knowing It

Layered Architecture, also known as N-Tier, is the most widely used architectural pattern in software development, organizing code into horizontal layers each with a distinct responsibility. The four core layers are Presentation, Application, Domain, and Infrastructure, where each layer communicates only with the one directly below it. This strict dependency rule means a Controller should call a Service, a Service should call a Repository, and never the other way around. The pattern's main strengths include conceptual simplicity, clear separation of concerns, and the ability to test or replace each layer independently. However, it can become rigid for vertical changes spanning all layers, and may produce redundant pass-through code when a Service layer adds no real logic.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Lilian Weng's blog post breaks down AI scaling laws and their real-world limits

AI researcher Lilian Weng published a detailed analysis titled 'Scaling Laws, Carefully' on her blog Lil'Log in June 2026, examining how model size, data volume, and compute collectively follow power-law relationships in large language model training. The post revisits the long-standing debate between the Kaplan scaling approach, which prioritized model size over data, and the Chinchilla findings, which showed that model parameters and training tokens should scale proportionally. Weng explains that the Chinchilla model, though four times smaller than DeepMind's Gopher, outperformed it by training on four times more tokens with the same compute budget. The post also addresses data-constrained scenarios, warning that repeatedly training on the same data yields diminishing returns and causes overfitting, especially in larger models. Weng cautions that scaling laws are empirical tools, not physical laws, and that small errors in curve-fitting can lead to vastly wrong predictions when extrapolating to expensive large-scale training runs.

0 comments Read more at DEV Community