Start Here
Two data-science perspectives on the neocloud industry. One playbook for ML / DS practitioners who run their work on neocloud infrastructure; one for the DS teams that build the analytics inside neocloud companies. Both perspectives matter; understanding both makes you better at either.
Why this guide
Data science intersects with the neocloud industry from two directions:
- From the outside in. A DS / ML practitioner running training, fine-tuning, evaluation, or research on neocloud compute. The decisions about providers, experiment design, and cost management are first-order data-science decisions.
- From the inside out. A DS team at a neocloud building pricing models, churn analysis, capacity forecasting, marketplace ranking, and anomaly detection. These are the analytical investments that make the company competitive.
This guide covers both. Mirroring the Data Engineering for Neoclouds guide structure, but with a DS / ML lens.
The two perspectives
Section B is the practitioner playbook. Five chapters: the DS workflow on GPU clouds, picking a provider for ML work, training and fine-tuning workflows, experiment tracking and reproducibility, and cost-aware experimentation.
Section C is the inside view. Five chapters: pricing models, churn and retention, capacity forecasting, marketplace ranking and matching, anomaly detection.
Section D synthesizes.
Who it's for
- ML / DS practitioners at AI companies whose workloads run on neocloud infrastructure.
- Data scientists at neocloud companies or interviewing for such roles.
- ML engineers who want to understand the analytics inside the infrastructure they depend on.
- Engineering and product leaders making provider or pricing decisions.
Reading order
Linear if you're new. Section B → Section C → Section D. If you only have time for the most-load-bearing chapters, read 02 (Picking a Provider) and 08 (Capacity Forecasting) — both flagged as critical.
Prerequisites
Assumes general DS / ML familiarity — supervised learning, evaluation, experimentation, basic statistics. Doesn't assume GPU expertise; the surrounding context helps. The Neocloud topic (separate guides) provides company background; this guide assumes you can reference it as needed.