ML Pipelines Flask DominoML

Building Pipelines with DominoML

Mar 15, 2026 • 4 min read • By Atharv Khare

Machine learning is inherently a DAG (Directed Acyclic Graph) of operations. DominoML turns this abstract concept into a tangible, visual drag-and-drop experience.

Writing boilerplate code for data ingestion, scaling, and model fitting is repetitive. I built DominoML to allow researchers and engineers to literally draw their pipelines.

The Node Architecture

Under the hood, every node is an atomic execution block in Flask that maintains its own state and dependencies. When you connect a Data Node to a Preprocessing Node, you're defining an edge in the computational graph.

Architecture: The canvas uses a custom JS renderer, while the execution engine relies on a topological sort to run nodes in the correct order.

Future Work

The next iteration involves distributed execution. Right now, pipelines run sequentially on the backend. Moving to a Celery-based worker queue will allow parallel execution of independent branches in the DAG.

... reads