Skip to content

KladML

Universal MLOps: Zero to Training in 60 Seconds

What is KladML?

KladML is a modular SDK for building production-ready machine learning pipelines. Unlike heavy MLOps frameworks, KladML gives you:

  • Universal Quickstart - Auto-detect data type, suggest pipeline, train in one command
  • Interface-based architecture - Swap backends without changing code
  • Local-first - No servers required, works offline with SQLite
  • High Performance - Powered by Polars (Data) and Torch Compile (Models)
  • Extensible - Register custom architectures, preprocessors, and evaluators
  • CLI included - Initialize projects, run experiments from terminal

Quick Install

# Core library
pip install kladml

# Full CLI with TUI
pip install "kladml[cli]"

Quick Start

Zero to Training in 60 Seconds

kladml quickstart --data my_data.csv

# Output:
# 📊 Analyzing data...
#    Data type: TABULAR (5 columns, 1000 rows)
#
# ? What task do you want to perform?
#   > Classification (detected 'label' column)
#
# 🔧 Selected: XGBoostClassifier + ClassificationEvaluator
# 🚀 Training...
# ✅ Complete! Results saved to data/projects/quickstart/run_001/

Supported Data Types

Data Type Auto-Detection Default Pipeline
TABULAR Numeric CSV/Parquet XGBoost (Coming Soon)
TIMESERIES Has datetime column Transformer
IMAGE Folder with JPG/PNG ResNet50
TEXT CSV with text columns BERT (Coming Soon)

Why KladML?

Feature KladML MLflow ClearML
Interface-based ✅ Pluggable ❌ Hardcoded ❌ Hardcoded
Server required ❌ No ⚠️ Optional ✅ Yes
Local-first ✅ SQLite default ✅ Yes ❌ No
Data Engine 🚀 Polars 🐢 Pandas 🐢 Pandas
Learning curve 🟢 Minutes 🟡 Days 🔴 Weeks
Universal Quickstart ✅ Yes ❌ No ❌ No

Documentation

  • 🚀 Getting Started — Install, configure, and run your first experiment
  • 🧠 Core Concepts — Understand interfaces, runners, and the architecture
  • 🏗️ Model Architecture — Deep dive into model contracts and design patterns
  • 🗺️ Roadmap — Planned features and what's coming next
  • 📦 CLI Reference — All available commands and options
  • 🚢 Deployment — Export and deploy to edge devices