Why MLArtisan Exists
I’m Alexey - a full-stack engineer who loves building tools that remove unnecessary complexity.
Today I’d like to share what I’ve been working on: MLArtisan.
The Pain That Started Everything
A few years ago, our small team tried to integrate ML into a product.
Not “deep research ML”, just practical things like predictions, clustering, or simple models per customer.
But here’s what actually happened:
Endless library research (many were Python-only, but we were on C#)
Dataset management became a project of its own
Experimentation was slow, disorganised, and hard to compare
The infrastructure setup took more time than the model itself
Deployment required writing custom services for each model type
Updating a model often meant updating the codebase
Customer-specific models added even more complexity
We spent more time on infrastructure than on machine learning.
At some point, I realised:
“To create an ML model, teams really need just four steps:
Schema > Data > Train > Predict.
Why is everything else so complicated?”
That became the seed of MLArtisan.
What MLArtisan Does
MLArtisan is a lightweight ML platform designed for developers, small teams, and indie builders who want ML without heavy infrastructure.
Define input/output schemas: This structures your data and lets you keep consistent versions.
Upload datasets: Version, filter, and prepare training data with minimal friction.
Train models for your case: Neural Networks, RNNs, K-Means, SVM, Naive Bayes, Logistic Regression, ONNX, and more. Each of them supports different types of Hyperparameters like learning rate, layers, batch size, clusters, etc. If your machine has a GPU, the app will use it automatically (self-hosted version).
Deploy instantly: Every model gets a prediction API. No pipelines, no extra services, no infrastructure setup.
There is a simple model registry:
Each model has versions
Last metrics stored
Copy models (e.g., "experiment" > "stable")
Improve one while another stays in production
More features, like tags and metric comparison, are planned.
Cloud or self-hosted: There’s a Docker image for teams who want full control: https://hub.docker.com/r/alexeymlartisan/mlartisan-selfhosted
Who It’s For
SaaS founders adding ML predictions
Indie hackers building data-driven features
Internal tools needing simple ML
Automation teams
People who want ML without learning TensorFlow/PyTorch
Anyone who wants to run it in the cloud OR entirely self-hosted
Basically, anyone who wants practical ML without rebuilding the same infrastructure from scratch.
Current State: Early Open Beta
The project is fully functional but still early.
You may find bugs, rough edges, or be slow in some places - and that’s totally expected.
You can try it here:
Cloud: https://mlartisan.com
Self-hosted: https://hub.docker.com/r/alexeymlartisan/mlartisan-selfhosted
Bugs/feedback: https://github.com/orgs/MLArtisan/discussions/categories/issues-bug-reports
Q&A: https://github.com/orgs/MLArtisan/discussions/categories/q-a
Ideas: https://github.com/orgs/MLArtisan/discussions/categories/ideas
I’m personally answering everything during the beta.
What I’m Looking For
Early testers
Honest feedback
Feature requests
Real-world use cases
Bug reports
People who want to integrate ML and tell me where things break
If you're building something and want to test MLArtisan, I’d love to hear from you.
You can comment here, DM me, or reach me at: contact@mlartisan.com
Thanks for reading!
I excited to get your thoughts!
