---
title: "S3 and object storage | Skills We Assign For | Azendo"
description: "Object storage for ML artefacts — layout, lifecycle, throughput, plus the Azendo roles assigned for it."
url: "https://azendo.co/skills/s3-and-object-storage/"
---

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

# S3 and object storage.

Object storage holds the large artefacts machine learning produces — datasets, model weights, checkpoints, feature files — durably and cheaply, addressed by key and accessed over HTTP.

## Where S3 and object storage fits on a long engagement.

Object storage is the substrate under most machine learning infrastructure. Training data, checkpoints and model artefacts all live there, and the durability and cost profile make it the right default for anything that is written once and read many times.

Key layout matters more than it appears. A prefix structure that supports the access patterns — listing a dataset version, retrieving a specific checkpoint — determines whether operations are efficient or require expensive listings, and it is very difficult to reorganise once terabytes are in place.

## What an assigned team does with S3 and object storage.

Training throughput is frequently limited by storage rather than compute. A GPU idle while waiting for data is expensive idle, and the fix is usually parallel reads, prefetching or a local cache rather than a larger instance.

Diagnosing that correctly requires measuring where the time actually goes, which is the kind of rigour held within an agreed [committed monthly capacity](https://azendo.co/pricing/).

## What we use S3 and object storage for.

* Artefacts stored durably and cheaply Datasets, checkpoints and weights on the right substrate for write-once read-many.
* Key layout matched to access Prefixes supporting the real operations, because reorganising later is impractical.
* Storage removed as the training bottleneck Prefetching and caching so the GPU is not idle waiting for data.

## How S3 and object storage capacity is assigned.

Artefact storage is assigned across AI and data capacity, with throughput measured before compute is increased.

## Roles we assign S3 and object storage for

* [ML Platform Engineer MLOps engineering](https://azendo.co/services/mlops-engineering/ml-platform-engineer/)

## Service lines it sits in

* [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 S3 and object storage 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/)