Скачать или смотреть AI workflows: golden paths for data scientists

At AI Dev 2025, I caught up with Diego Braa, CTO & co-founder at Krateo, to unpack their open-source PlatformOps approach: using Kubernetes + Operators to automate heterogeneous infrastructure (on-prem, cloud, legacy via APIs), expose it via an Internal Developer Portal, and keep spend in check with FinOps-first guardrails (think “lights-off on weekends” and right-sizing by default). We also touch on why data scientists are “the new developers,” and how k8s can be the control plane for modern AI workloads. If you’re moving from classic config management (Ansible/Chef/Puppet/Terraform) toward platform engineering, this one’s for you.

Chapters
00:00 Intro & guest
00:35 Project origin (2021, solving automation fragmentation)
01:40 Why Kubernetes Operators for infrastructure automation
02:30 Turning automation into an Internal Developer Portal
03:05 FinOps features: cost controls, schedules, carbon-aware toggles
03:55 AI workflows: golden paths for data scientists
04:40 Looking ahead to Kubernetes 1.35
05:05 Get involved: GitHub org, self-service labs, docs
05:45 Connect with Diego & Cateo
06:00 Outro

Mentioned
Kubernetes, Operators, PlatformOps
Internal Developer Portal (IDP)
FinOps: cost optimization, off-hours shutdowns, right-sizing
Hybrid & legacy via API control

If you found this useful, hit 👍 and subscribe for more practical platform engineering + AI build tips.