MLOps

Ray43.9k★

Ray lets you scale Python code from your laptop to a cluster by adding a decorator to your functions. No rewriting your code, no learning a new framework. It handles distributed computing, model training (Ray Train), hyperparameter tuning (Ray Tune), model serving (Ray Serve), and reinforcement learning (RLlib). Fully free under Apache 2.0. The core engine and all libraries are open source with no feature gating. Anyscale (the company behind Ray) offers a managed platform, but self-hosting the full stack costs $0. The catch: Ray's "just add a decorator" marketing undersells the complexity. Distributed computing is hard, and Ray doesn't eliminate that; it manages it. Debugging distributed tasks, understanding memory management across workers, and tuning cluster resources requires real expertise. For single-machine ML, you don't need Ray. It earns its place when you need to scale beyond one box or orchestrate complex ML pipelines.

Gradio43.6k★

No JavaScript, no frontend knowledge required. Apache 2.0, Python. Backed by Hugging Face (they acquired the company). The API is dead simple: define your function, specify input/output types, and Gradio generates the web interface. Supports text, images, audio, video, files, dataframes, and custom components. Share instantly via a temporary public URL or deploy permanently on Hugging Face Spaces for free. Fully free. No paid tier for the library itself. Hugging Face Spaces offers free hosting with some compute limits; paid Spaces start at $7/month for dedicated hardware. Solo ML engineers: this is how you demo your work. Build a model, wrap it in Gradio, share the link. Takes 10 minutes. Small teams: use it for internal tools and stakeholder demos. Medium teams: build it into your ML workflow for model evaluation interfaces. The catch: Gradio is built for demos and internal tools, not production apps. The generated UIs are functional but not customizable enough for customer-facing products. Performance degrades with concurrent users; it's running your Python function synchronously by default. For production ML serving, use a proper API (FastAPI + frontend) instead of a Gradio wrapper.

Label Studio28.3k★

Draw bounding boxes on images, highlight entities in text, transcribe audio, or create custom labeling interfaces with a template system. It's the Swiss Army knife of data labeling. The community edition is free under Apache 2.0. You get multi-user support, a project system, customizable labeling interfaces, and import/export in every format (COCO, YOLO, spaCy, etc.). Self-host it with Docker and you're labeling data in 10 minutes. The catch: the free version is single-node only. Label Studio Enterprise (now called HumanSignal) adds team management, active learning, model-assisted labeling, RBAC, SSO, and analytics, but pricing requires a sales call. The community edition handles small-to-medium labeling projects well, but once you have 5+ annotators who need quality control and agreement metrics, you'll feel the feature gap.

MLflow28.0k★

MLflow tracks the entire machine learning lifecycle: experiments, parameters, metrics, model versions, and deployment. It is version control for your ML work, where every training run and every result gets logged and compared. Apache 2.0, backed by Databricks, free to self-host. MLflow 3 widened the scope hard toward LLMs and agents. Tracing captures prompts, retrievals, tool calls, and responses so you can debug an agent that went sideways. Evaluation runs LLM-as-judge metrics instead of you eyeballing outputs. There is a prompt registry for versioning prompts, and OpenTelemetry compatibility so the traces are not locked to one vendor. Self-hosting is still a pip install, a server process, and pointing your scripts at it: SQLite for solo work, Postgres once more than one person writes to it, Docker images if you would rather not think about it. Solo ML engineers: self-host, free, done. Teams of two to ten: self-host with Postgres and accept a couple of hours a month keeping the backing store and artifact storage healthy. Ten to fifty is where RBAC and audit trails start to matter and the managed option earns a look. Databricks sells managed MLflow with integrated compute and governance, billed in Databricks Units rather than a flat seat price, so cost tracks your cloud and your usage rather than your headcount. The catch is that MLflow tracks everything and does not do everything. You still need compute, a feature store, and production monitoring from somewhere else. The Databricks integration runs deep enough that newer features tend to land on managed first, and the pivot toward GenAI means the classic ML surface gets less of the spotlight than it used to.

Weights & Biases11.3k★

Weights & Biases logs every detail of your ML experiments (parameters, metrics, outputs, hardware usage) and gives you dashboards to compare runs side by side. It's a lab notebook for machine learning that actually stays organized. MIT license, Python. Two lines of code to integrate: `wandb.init` and `wandb.log`. Works with PyTorch, TensorFlow, Keras, Hugging Face, and basically every ML framework. The visualization is where it shines: interactive charts, parameter importance plots, and run comparisons that MLflow's UI can't touch. Free tier: unlimited personal projects, 100GB storage. That's generous for solo work. Team plan starts at $50/user/month: RBAC, audit logs, team dashboards. Self-hosting exists (W&B Server) but it's enterprise-only. No free self-hosted option. You're on their cloud or you're paying for a private deployment. Solo ML engineers: the free tier is excellent. Use it. Small teams (2-10): $50/user/month adds up fast ($500/mo for 10 people). Compare against self-hosted MLflow at $0. Growing teams: the collaboration features justify the cost if experiment tracking is critical to your workflow. The catch: no free self-hosted path. Once your team grows, you're locked into per-seat pricing or an enterprise contract. MLflow gives you the same core tracking for free if you're willing to run it yourself.

skypilot10.6k★

SkyPilot runs your AI training jobs on whichever cloud has GPUs available. Write the job once, and it figures out where to land it: your Kubernetes cluster, your Slurm cluster, AWS, GCP, Azure, RunPod, Lambda, 20+ other clouds. It handles spot instance failover, queue management, and auto-cleanup of idle resources. Apache 2.0, install with pip. Self-hosted reality is straightforward for the user side. Pip install, configure cloud credentials, write a YAML spec. The infra-team side is heavier: for shared clusters with multi-tenancy, gang scheduling, and team resource quotas, you're running their API server and tuning it. Most solo users skip that and just point SkyPilot at their existing cloud accounts. Solo ML engineers chasing GPU availability across clouds: this is the move. Small teams sharing a Kubernetes cluster or split across GCP and AWS: same. Large ML platforms at companies like Shopify already use it. Enterprise features (SSO, RBAC, advanced governance) push you toward the paid SkyPilot offering, but the core is unlimited. The mental model has a learning curve. SkyPilot speaks YAML and CLI fluently. For a click-to-launch UI for non-engineers, you'll need to build that on top.

Aim6.3k★

Aim is an open source experiment tracker that goes head-to-head with Weights & Biases on features and beats it on data ownership. Track training runs, log parameters, compare metrics, and run aggregations across thousands of experiments through a web UI that doesn't slow down the way TensorBoard does past a few hundred runs. The team claims smooth UI behavior at 10,000+ runs, which holds up in practice. Install is `pip install aim` and `aim up` to launch the UI. The Python SDK plugs into PyTorch Lightning, Hugging Face, Keras, XGBoost, and the usual suspects. Built-in converters import existing TensorBoard, MLflow, and W&B logs, which makes migrating cheaper than starting fresh. A remote tracking server lets a team share one centralized Aim instance. Solo researchers: install it. There's no friction. Small teams: run a shared server and stop paying for W&B Personal seats. Large teams: Aim handles the scale, and you keep the data on your infrastructure. The AimStack team offers enterprise support if you want a contract behind the deployment. The catch: Aim is excellent at run comparison, less full-featured than MLflow at the broader lifecycle (model registry, deployment hooks). If you want a single tool that tracks experiments AND serves models, MLflow is more end-to-end. If experiment comparison is the core need, Aim is the better UI.