---
title: "ONNX | Skills We Assign For | Azendo"
description: "ONNX for model portability — export, runtime optimisation, plus the Azendo roles assigned for it."
url: "https://azendo.co/skills/onnx/"
---

[Skills](https://azendo.co/skills/) MLOps and model delivery 

# ONNX.

ONNX is an open format for representing machine learning models, allowing a model trained in one framework to be exported and run in another runtime, including optimised engines and constrained environments.

## Where ONNX fits on a long engagement.

Separating training from serving is the architectural value. A model can be trained in PyTorch and served through a runtime optimised for inference — often faster, with fewer dependencies, and available on platforms where the training framework is not.

Export is not always clean. Models using unsupported operators, dynamic control flow or custom layers may fail to export or export with subtly different behaviour. Verifying numerical equivalence between the original and the converted model is a necessary step, not a formality.

## What an assigned team does with ONNX.

The deployment benefits are real where the target is constrained. Edge devices, mobile and environments where installing a full training framework is impractical are all cases where a portable format is what makes deployment possible at all.

Establishing the target constraints before choosing the training approach is basic sequencing, part of the scoping described in [how an assignment runs](https://azendo.co/how-it-works/).

## What we use ONNX for.

* Training and serving decoupled A model trained in one framework and served through an optimised runtime.
* Equivalence verified after export Numerical output compared, because conversion can change behaviour subtly.
* Deployment to constrained targets Edge and mobile environments where a full framework cannot be installed.

## How ONNX capacity is assigned.

Model portability work is assigned inside AI capacity, with post-export equivalence testing treated as a required step.

## Roles we assign ONNX for

* [Computer Vision Engineer AI engineering](https://azendo.co/services/ai-engineering/computer-vision-engineer/)
* [Model Deployment Engineer MLOps engineering](https://azendo.co/services/mlops-engineering/model-deployment-engineer/)

## Service lines it sits in

* [AI engineering](https://azendo.co/services/ai-engineering/)
* [MLOps engineering](https://azendo.co/services/mlops-engineering/)

Capacity is agreed as a committed monthly capacity across a discipline, not per skill.

## Related in mlops and model delivery

* [TensorRT — skill we assign for](https://azendo.co/skills/tensorrt/)
* [Model registry — skill we assign for](https://azendo.co/skills/model-registry/)
* [Weights & Biases — skill we assign for](https://azendo.co/skills/weights-and-biases/)
* [Comet — skill we assign for](https://azendo.co/skills/comet/)
* [Neptune.ai — skill we assign for](https://azendo.co/skills/neptune-ai/)
* [DVC — skill we assign for](https://azendo.co/skills/dvc/)
* [Feast — skill we assign for](https://azendo.co/skills/feast/)
* [feature stores — skill we assign for](https://azendo.co/skills/feature-stores/)

## Tell us what your roadmap needs ONNX for.

A service delivery manager replies with the disciplines we would assign, the monthly capacity and what the first month looks like.

[All skills we assign for](https://azendo.co/skills/)