Pulumi ads

The Pulumi data team ran into a problem many data teams are hitting right now. Letting an LLM generate SQL for analytics sounds great until the answers look right but are wrong. Missing joins. Bad filters. Incorrect assumptions. That’s worse than no answer at all. What actually worked was treating the LLM as an interface, not the source of truth, and constraining it behind a semantic layer on top of Snowflake that defines metrics and relationships. If you’re building analytics agents and care about trust and reliability, this breaks down the real tradeoffs we had to make: https://hubs.ly/Q042HsGj0 #DataEngineering #AIagents #LLMs #snowflake

AI teams are building across #AWS Bedrock, #Azure OpenAI, GCP Vertex AI, and specialized GPU providers. Managing this complexity shouldn't require learning proprietary configuration languages or getting locked into a single vendor's tooling. ▪️Supabase manages 80,000 resources across 16 AWS regions in TypeScript. The same language they use for their application services. ▪️Snowflake runs 100,000+ daily deployments across three major clouds. ▪️BMW Group enables 15,000 developers to access self-service infrastructure. Infrastructure that scales with superintelligence workloads while remaining accessible to every ML team. Built with code, not configuration languages. Learn more at https://hubs.ly/Q03-0shd0 #AIinfrastructure #PlatformEngineering #MLops

Pulumi's IaC platform will now support Terraform/OpenTofu clients, our own open source IaC tool will speak HCL, and we're making it as easy as ever before to say yes by letting you use HashiCorp purchased credits for Pulumi usage. These new capabilities are in private beta and will go public in Q1'2026. We know that many folks are thinking about 2026 as we head into the holidays and infrastructure is as important as ever, especially with AI in the air. We've worked with thousands of cloud innovators and we regularly hear that they want to go all in on Pulumi but that they have lots of investment into Terraform holding them back. We also hear that sometimes there are folks on the team more comfortable with HCL which they have considerable muscle memory with, even if the central platform team prefers general purpose languages. Now it's not an either-or question: You can standardize on the Pulumi platform and accommodate the full spectrum of IaC tools and languages. You benefit from the entire platform — automation, Neo, our infrastructure engineering agent, governance and insights, our IDP, and more — regardless of your choice. We also continue to hear frustration with HashiCorp, an IBM Company, and how they turned their back on open source, are hiking prices unreasonably during renewals, dropped support for CDKTF with zero days advance notice, are confusing everybody with free tier retirement (and unretirement), and more. In contrast, our number one company value is "when the customer is successful, we are too," and we stand strongly by our open source heritage. If you're reconsidering the IaC platform you're betting on for the next decade, we would be honored to earn your business. Check out the blog to learn more, DM me, or shoot a note to [email protected] — we're ready to roll up sleeves. https://lnkd.in/g23mqfU3

The gap between AI research and production infrastructure is real. Research teams iterate fast with PyTorch and Ray. Production teams need #Kubernetes orchestration and Airflow pipelines. Most companies rewrite infrastructure at every stage. Engineers building superintelligence systems at leading AI companies confirm there's a better path: manage your entire ML infrastructure in code your teams already know. From research experiments on a single GPU to thousand-node training clusters at a massive scale. Infrastructure is written in general-purpose programming languages. No proprietary configuration syntax. No rewrites between environments. Code speaks to code. Discover Superintelligence Infrastructure: https://lnkd.in/eK8GWehQ/?utm_medium=paid-social&utm_source=linkedin&utm_campaign=FY2026_Superintelligence-Infrastructure&utm_id=804194246&hsa_acc=508076010&hsa_cam=804194246&hsa_grp=488289906&hsa_ad=1002735786&hsa_net=linkedin&hsa_ver=3 #MLops #PlatformEngineering #DevOps #InfrastructureAsCode

CI/CD pipelines still rely on long lived credentials that are stored as secrets, copied across systems, and rarely rotated as often as they should be. Pulumi supports native #OIDC token exchange for the CLI to change that model. Instead of storing personal or organization access tokens, CI/CD environments can authenticate to #Pulumi Cloud using the OIDC credentials they already issue. The CLI exchanges that identity for a short lived Pulumi Cloud token at runtime. This works with GitHub Actions, GitLab CI, and #Kubernetes environments such as EKS, GKE, and AKS. Access can be scoped to specific users or teams, with no new secrets introduced into your pipelines. No token sprawl. No rotation overhead. Just a simpler and more secure way to authenticate infrastructure automation. Read more at https://lnkd.in/gENFAgZz #DevSecOps #CICD #CloudSecurity

We are seeing an explosion of AI infrastructure, from NVIDIA and frontier labs to high tech innovators like Snowflake to enterprises who are now fine-tuning their own models on their unique datasets. As the industry collectively spends $100Bs annually on scaling these architectures, new needs have emerged, including highly dynamic reinforcement learning approaches, GPU fungibility, and overnight ubiquity of neo-clouds. At this scale, humans often become the bottleneck for scaling, demanding that we use AI itself to automate AI infrastructure. This then speeds up training, serving, and iteration, making the models even better. At Pulumi, we call this the superintelligence flywheel, and you can read more here: https://lnkd.in/gq_ZZz58 #superintelligence #pulumi