Decision Layer

MLX

From Raw Data to
Production ML — Accelerated

End-to-end machine learning and time series forecasting with data exploration, automated feature engineering, intelligent model training, and Databricks integration — all orchestrated from one platform.

6 ML Models|8 Forecasting Models|AutoML + Agentic AutoML|Databricks Native|MLflow Tracking

Start Exploring Data

Your ML Journey

Five Levels of ML Maturity

Most organizations are stuck at Level 1 or 2. MLX provides the platform to reach Level 5.

Ad-Hoc Analysis

Where most teams start

Data lives in spreadsheets and notebooks. Analysis is manual, inconsistent, and impossible to reproduce. Every question requires starting from scratch.

Spreadsheets and ad-hoc scripts

No standardized data profiling

Results vary between analysts

MLX enables this with:Data Exploration

The Right Approach

Manual, Automated, or Agentic?

All three approaches are available for both classification/regression and time series forecasting.

Manual Training

Full control, full precision

Best When

You know which algorithm fits your problem

You need full control over hyperparameters

You want transparent, interpretable models

You Get

Precision-tuned models for known problems

Full transparency in every decision

Fast iteration for experienced practitioners

Supports classification, regression, and forecasting

Higher effort, maximum control

Train a Model

AutoML

Let the machine search

Best When

You have clean, prepared data

You want to compare multiple algorithms

You need a strong baseline quickly

You Get

Breadth of comparison across 10+ algorithms

Auto-selection based on your metrics

Generated performance reports and leaderboards

Available for both ML and time series forecasting

Medium effort, broad coverage

Launch AutoML

Agentic AutoML

AI drives end-to-end

Best When

You have raw, unprocessed data

You want end-to-end automation

You need the agent to explore and decide

You Get

Autonomous data exploration and feature engineering

Agent-driven model selection and optimization

Human oversight at critical decision points

Supports classification, regression, and forecasting pipelines

Lowest effort, highest automation

Start Agentic Job

End-to-End Workflow

From Data Upload to Deployed Model

Ingest

CSV upload or Databricks import. Data lands in MinIO with automatic type detection.

Explore

Statistics, correlations, distributions, missing value analysis, and time series decomposition — all automated.

Engineer

Feature importance, selection, automated generation with configurable primitives.

Train

Manual training, AutoML comparison, or Agentic end-to-end — for classification, regression, and time series forecasting.

Evaluate

Metrics dashboard, model comparison, leaderboard. Every experiment tracked in MLflow.

Deploy

Model registry, ONNX export, version management. Production-ready artifacts.

Ingest

CSV upload or Databricks import. Data lands in MinIO with automatic type detection.

Explore

Statistics, correlations, distributions, missing value analysis, and time series decomposition — all automated.

Engineer

Feature importance, selection, automated generation with configurable primitives.

Train

Manual training, AutoML comparison, or Agentic end-to-end — for classification, regression, and time series forecasting.

Evaluate

Metrics dashboard, model comparison, leaderboard. Every experiment tracked in MLflow.

Deploy

Model registry, ONNX export, version management. Production-ready artifacts.

Enterprise Integrations

Built for Your ML Stack

Databricks Unity Catalog

Connect your lakehouse

Connect Databricks workspaces with API tokens

Browse Unity Catalog hierarchy directly

Import tables into MLX for exploration

Zero data movement — read-only access

Connect Databricks

MLflow Experiment Tracking

Every run, every metric

Automatic experiment logging for every training job

Model versioning with full lineage

Compare runs side-by-side on metrics

View Experiments

The Transformation

What Changes with MLX

Without MLX

With MLX

Jupyter notebooks — inconsistent, not reproducible

Standardized exploration with statistics, correlations, and distributions

Feature engineering requires deep pandas expertise

Automated feature generation with configurable primitives

Model training one algorithm at a time

AutoML compares 10+ algorithms in a single job

No systematic model comparison

Side-by-side metric comparison and leaderboard

Manual MLflow setup and tracking

Automatic experiment logging and model registry

Time series forecasting requires specialized statistical knowledge

Guided forecasting with automated decomposition, model selection, and evaluation

Weeks to first production model

Hours to first production model

Under the Hood

The ML Pipeline

From raw dataset to production model — explore, engineer features, and train with manual control, AutoML, or AI agents.

Dataset

CSV · Databricks

upload

ML Pipeline

EXPLORE

Data Analysis

pandas · scipy

features

ENGINEER

Feature Engineering

Importance · Selection · PCA

engineered data

TRAIN

Model Training

scikit-learn · XGBoost

ONNX model

Model

ONNX · MLflow

training mode

Training Approaches

Manual

Full control

AutoML

Auto search

Agentic

LLM agents

DatasetCSV · Databricks

upload

ML Pipeline

EXPLORE

Data Analysis

pandas · scipy

features

ENGINEER

Feature Engineering

Importance · Selection · PCA

engineered data

TRAIN

Model Training

scikit-learn · XGBoost

training mode

Training Approaches

Manual

Full control

AutoML

Auto search

Agentic

LLM agents

ONNX model

ModelONNX · MLflow

Infrastructure Layer

PostgreSQLMetadata & results

MinIODataset & model storage

MLflowExperiment tracking

OpenTelemetryObservability

Ecosystem

The DecisionOS Ecosystem

MLX is the ML training layer — drawing on KnowledgeX for feature context, DataX for data quality, ModelsX for inference, MonitoringX for observability, and SemanticX for domain terminology.