DataOps & MLOps Engineer
Pubblicato il 27-09-2026 - ALLSIDES in Bressanone
About ALLSIDES
Si ricordi di controllare il suo CV prima di candidarsi. Inoltre, si assicuri di leggere tutti i requisiti relativi a questo ruolo.
ALLSIDES is redefining how the world experiences 3D content. We combine physically accurate scanning and generative AI to power content creation workflows for e-commerce, virtual environments, and immersive experiences. Our clients include global brands like adidas, Meta, Amazon, and Zalando.
We operate a rapidly scaling photorealistic 3D scanning operation, capturing tens of thousands of assets annually while training next‑generation AI models. As an NVIDIA Inception member, we collaborate with leading research institutions and actively participate in top‑tier conferences in 3D computer vision and AI.
More info: | Overview
We're looking for a DataOps & MLOps Engineer to build the infrastructure that powers our data and ML workflows. You'll focus on data storage and movement, dataset versioning, ML pipeline automation, experiment tracking, and ensuring reproducibility across our 3D reconstruction and training workloads.
Main Responsibilities
-
- Design and manage data storage systems for large datasets (multi‑TB image data, 3D assets, training data)
- Build efficient data access patterns and movement strategies for distributed training and experimentation
- Implement dataset versioning and lineage tracking for reproducibility
- Set up and maintain experiment tracking and model registry infrastructure (MLflow, Weights & Biases)
- Build ML pipelines for data preprocessing, training, validation, and model registration (Kubeflow, Airflow, Prefect)
- Support distributed training workflows across multi‑GPU clusters (PyTorch Distributed, Horovod, Ray)
- Profile and optimize training pipelines: data loading bottlenecks, batch sizing, GPU memory utilization
- Ensure reproducibility of experiments: environment pinning, data versioning, artifact management
- Manage artifact storage and distribution (Docker registries, model reg