← All organizations

Machine learning pipeline development

Sematic

Sematic developed an open-source ML pipeline platform for machine learning engineers and data scientists. Its Python framework connects data processing, model training, and other business logic into workflows that run locally, on cloud virtual machines, or on Kubernetes. A web dashboard tracks pipeline inputs, outputs, and artifacts, helping developers inspect their work. Individual steps can use different compute resources, including CPUs, GPUs, and Spark clusters.

Co-founder Emmanuel Turlay previously led ML platform work at Cruise, and Sematic’s design drew on experience building self-driving machine learning infrastructure. Its technical approach supports nested pipelines, Python-defined loops and conditional branches, runtime type checking, and cached steps. Developers can run the same pipeline code on a laptop and a Kubernetes cluster, connecting local experimentation with cloud execution.

The company pivoted to Airtrain in summer 2023, applying its orchestration technology to no-code LLM evaluation and fine-tuning. Airtrain targeted developers building AI products; in February 2024, Turlay described a few hundred self-service users and an inference API billed per token. As of August 2026, Y Combinator lists Airtrain AI as inactive; that classification does not establish the maintenance status of Sematic’s open-source project.

sematic.dev

1 talk

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Messages from the stage

Comparing models with cost in view

Turlay contrasts GPT-4 with the less costly Flan-T5 and introduces Airtrain for dataset-based comparisons involving models such as Llama 2 and Falcon.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-28