Hi! Is anyone here using AI in their platform engineering? I am registered for the upcoming webinar and hopefully that gives me some clarity but I do not have an easy time seeing what I would actually use AI to do. Based on the findings in the new report - I am not alone in not using or seeing value in AI for platform engineering.

Hey Toby! Good question @Natan Yellin. He’s pretty actively working on the AI stuff here.

I want to experiment next year with AI and ways to improve productivity. I see potential leverage in improving quality of tests or mocks. Haven't received an invite to the webinar, would you mind sharing?

I have friends working on AI as a guardrail for Platform Engineering, i.e. review IaC and then make updates to ensure data leakage does not happen. Gombac.ai

in a narrower domain in this field, for IaC we have seen a lot of traction for self service capabilities (App folks using ready to use templates to deploy foundation infra for their needs) AND auto app architecture mapping to reduce the workload on engineering teams to define future state. I have seen use case for ci/cd standardization and kpi generation using AI (not necessarily ready made models but unsupervised learning). Disclaimer - its our product in a very narrow field for IaC (generate code in any language of your choice AND more importantly visualization to confirm, fine tune, self service consumption etc. in qmcloud.io).

@Gregory Boch https://platformengineering.org/events/how-palo-alto-networks-is-transforming-platform-eng-with-ai-agents

A few. 1) if you compose your desired state using our canvas, you can genrate code in tf, cf or arm (2) or if you use enter a drscription of what you want, a visualization is generated that allows you to validate the config, dependencies etc and finw tune.. without the validation and visualization I have seen folks struggle with what copilot gives you. Besides that we can deploy if you want. We use pulumi in the backend but that's transparent.

Toby Kam, it works well. It utilizes deterministic AI for this purpose. IIRC there's a 2 week trial.

Hi @Toby Kam, re your original question, it would be helpful to know more what your team does on a day to day basis to answer!

I'm engineering platforms for AI... I guess eventually they will start engineering themselves?

Hi @Natan Yellin. I do not currently have a team. In the past, I was in charge of CI/CD process.

Got it, thanks. In the CI/CD area, I’ve seen a few companies do automatic AI analysis of failed Jenkins jobs, so that devs can fix broken builds faster

a blog post we have written about ai/infra: https://www.anyshift.io/blog/navigating-ai-in-your-infrastructure-dos-don-ts-and-why-it-matters

Better to share some context and takeaways from the blog than just sharing it outright @Roxane Fischer

your right. Key takways: • why genai for IaC is not good enough in some cass (-> need some context / guidelines / where to fetch relevant information) • the distinction with what researchers call "synthesis ai" : gen ai is really good to find some comon patterns, read huge amount of data -> root cause analysi • how to improve IaC generation with graph retrieval (your infra in a graph)