CIQ

Join CIQ and Phinite.ai at CIQ headquarters during LA Tech Week for “Sovereign AI: From GPU to Governed Production.” See what it takes to move private AI beyond the pilot through live demonstrations of: • An open-weight model served on CIQ’s AI-ready stack inside the customer environment • Phinite agents running on that model with complete traces, guardrails, and human approvals • A model upgrade tested, promoted, and rolled back with a complete audit trail This event brings together leaders from startups, venture capital, and the enterprise, including CTOs, CISOs, Heads of AI and Platform, and anyone responsible for moving private AI into production without giving up control. Saturday, October 17 1:00 to 6:00 PM CIQ HQ, 5788 W Adams Blvd, Los Angeles Seats are limited. Register: https://lnkd.in/et2Hg9Ck #LATechWeek #SovereignAI #AIInfrastructure #EnterpriseAI

Welcome to CIQ, Amr Moussa!

Welcome to CIQ, Meaghan Parsons!

Scientists should not need a second career in computer science to use HPC. That belief has guided David Godlove from neuroscience research at Vanderbilt and NIH to supporting the Biowulf supercomputer, contributing to Singularity/Apptainer, and helping shape Fuzzball at CIQ. In our latest employee spotlight, David shares how each stage of his career has focused on the same goal: removing technical roadblocks so other people can solve important problems. Today, he brings the perspectives of a scientist, HPC user, software developer, educator, and solutions architect to his role as technical product writer for Fuzzball. His work includes testing new capabilities, bringing user feedback to engineering, and helping make features such as Fuzzball model serving approachable for researchers and enterprise AI teams. Read David’s story: https://lnkd.in/eqCHhjMt #Fuzzball #HPC #AIInfrastructure #EmployeeSpotlight

Updating Linux is a solved problem. Right up until you have a hundred servers. Frontier distributions get you new packages fast, and every so often something bad rides along with them. Legacy Enterprise Linux trades that for stability, but the vendor decides what ships and when, and your custom packages are left to fend for themselves. Brian Rieb breaks down where that model creaks at fleet scale, and how CIQ Enterprise Linux Manager (ELM) puts the decision back with your team. Pin production to what you tested. Promote when you're ready. The post includes a full walkthrough video, too. Read the blog post and watch the demo: https://lnkd.in/ewk-z9pG #EnterpriseLinux #LinuxManagement #RLCPro #SysAdmin

“We’re just upgrading Linux this weekend.” That sentence has started a lot of projects that ran for months. On October 22, CIQ’s Brady Dibble and Brian Rieb will sort out what separates an update, an upgrade, a conversion, and a migration, and what each one actually asks of your team. In this 30-minute session, they’ll discuss: • Why moving between binary compatible distributions is a conversion • Where containers remove the operating system from the equation, and where the host still matters Join the premiere and bring your questions to the chat, or register to receive the recording. Thursday, October 22 11:00 AM PT / 2:00 PM ET Register: https://lnkd.in/e2gm9_W6 #EnterpriseLinux #LinuxMigration #Linux #RLCPro

SingularityCE reaches end of life. Apptainer carries on. On October 25, Sylabs will end SingularityCE, SingularityPRO, and Singularity Enterprise. The open source container project it helped shape continues with Apptainer. Jonathon Anderson reflects on the history of Singularity, from its beginnings as Gregory Kurtzer’s project at Lawrence Berkeley National Laboratory to its evolution into the Linux Foundation project now known as Apptainer. Jonathon also thanks Sylabs for releasing previously commercial functionality as open source, giving those capabilities an opportunity to live on through the community. For current Singularity users, the path forward should feel familiar. Apptainer runs the same SIF images and continues to provide the singularity command for compatibility. For teams that need more than the runtime, Apptainer Pro will add long-term support, compatibility validation, execution records, and cluster image distribution, with early customers now being accepted. Read Jonathon’s perspective: https://lnkd.in/eyDNdXyG #Apptainer #Singularity #HPC #OpenSource

CIQ has announced Apptainer Pro, an upcoming commercially supported edition of Apptainer for organizations running AI and HPC containers at scale. Apptainer Pro will provide: • Designated long-term support releases with backported security and critical bug fixes • Preflight checks for GPU, CUDA or ROCm, and MPI compatibility • Execution records showing which image, user, runtime, and host were involved in each workload • Reusable OCI to SIF conversions, image prestaging, and managed node local caching • Local mirrors and signed offline updates for air gapped and sovereign environments Because Apptainer Pro builds on open source Apptainer, Singularity users can keep their existing SIF images, workflows, and the singularity command. CIQ is now accepting early customers, with priority for organizations moving from Singularity. Learn more: https://lnkd.in/e_ahg-xk #Apptainer #HPC #AIInfrastructure #EnterpriseLinux

One command puts a Hugging Face model on your own GPUs, served and ready for traffic. Jonathon Anderson walks through how Fuzzball does it by assembling the serving stack: a replica pool, a built-in LiteLLM gateway, and a stable OpenAI compatible endpoint. Jonathon also covers: • Eleven validated model presets with the GPU fit, context size, and vLLM arguments already configured • Pools that scale with demand and can idle at zero • OpenCode and hermes agent automatically discovering model gateways • Persistent node local volumes for environments without shared storage • Changes administrators should review before upgrading Learn to serve a model on your own infrastructure with Fuzzball. Read the blog post: https://lnkd.in/eZZB5kq2 #Fuzzball #AIInfrastructure #LLM #SovereignAI

Private AI should fit the infrastructure you already run. The latest Fuzzball release includes a governed platform to run open-weight models and agents on your own infrastructure. The release brings together: • Eleven validated open weight models • Automatic discovery between in-cluster agents and running models • Inference across NVIDIA and AMD GPU fleets • And still offers portability across on prem and major clouds With Fuzzball, move AI projects from pilot to production without assembling a new infrastructure stack or giving up control of models, data, or compute. Read the the Techstrong.ai article: https://lnkd.in/eidNqTDf #Fuzzball #SovereignAI #AIInfrastructure #OpenSourceAI