Nebius AI

🇦🇹 Nebius AI at ICML: whom to meet and where to find us One of the most anticipated events in our field, the [ICML] Int'l Conference on Machine Learning, will begin on Monday, July 22. We, of course, couldn’t miss it. Here are the members of our team who will be at ICML in Vienna: - Boris Yangel, Head of NLP - Sergey Polezhaev, ML Engineer - Levon Sarkisian, Cloud Solutions Architect Team Leader - Aleksandr Patrushev, Senior Product Manager on ML/AI - Simone Lonchiar, Sales Development Representative - Anna Peshekhonova, Head of Growth Marketing - Alina Vasilchenkova, ML Community & Events Manager Let’s meet and discuss your interests! Also, be sure to come say hi at our booth 202. We are in for a very productive conference. Next week, we will tell you more about our participation. #ICML #research #MLconferences #papers

⭐️ Introducing Managed Service for Apache Spark Request access if you’d like to process large-scale datasets using Apache Spark in the Nebius infrastructure: https://lnkd.in/eV3_Dg3w. Currently, the service is provided free of charge and is at the Preview stage. Here’s what we offer as part of it: - Low upkeep Focus on building queries, not infrastructure. We maintain and optimize Spark for you, so you can concentrate on your data processing tasks. - Big data processing Effortlessly handle large-scale data jobs. Easily manage jobs for calculations on large amounts of data during your dataset preparation. - Easy scaling Scale in seconds. Add new Spark clusters or increase their capacity quickly, with configurable resource usage limits to match your needs. - Serverless solution Resources are spent flexibly only on what you need. Control your consumption, which includes running jobs, active sessions, and configured History Server. - Diverse types of access Interact with Spark from the environment where you are most comfortable. The service supports various interfaces, from CLI and UI to IDE and Jupyter Notebooks. Learn about use cases where Spark is essential and who controls what in a managed model: https://lnkd.in/ehBjvt4i #Spark #datapreparation #dataprocessing #datasets

Today is #AIAppreciationDay. Our marketing team wants to thank AI tools and their creators for the AI magic that amplifies our productivity. Thanks a lot, dear Recraft, Grammarly, Runway, ChatGPT, Adobe and Perplexity! A special thank-you goes to Recraft for illustrating our postcards 😊.

Hardcore CUDA Hackathon: the winners are in! 🏆🏆🏆 That's a wrap for the hackathon at AGI House in San Francisco, which we sponsored with our H100 GPUs! 43 participants had been using Nebius infrastructure today. David, Evan, another Evan and Kevin are the winners with their very interesting steriOMG project for Apple Vision Pro. Our congrats go to: David Heineman — github.com/davidheineman Evan Rusmisel — github.com/enva2712 Evan Pierce — github.com/mrthinger Kevin Tran — github.com/hadondish Huge thanks to Ash Vardanian, Jeremy Nixon, Rohan Pandey and Kyle Morris for organizing! #winners #hackathon #CUDA #AGIHouse

Meet our own Idan Belisha, a Cloud Solution Architect, explaining to the participants of the CUDA hackathon how to start using H100s. We take pride in our architect team, and it’s great to have Idan with us here in the Bay Area. #CSA #GPUcloud #architects #hackthons

🏞️ Today, we’re at the Hardcore CUDA Hackathon in San Francisco Good night, Europe, and good day, America! We just kicked off the hackathon with approximately 100 participants, to whom we are providing NVIDIA H100 GPUs. Our quAIck-quAIck is also around. #H100 #GPUs #hackathons #quAIck

🏙️ Introducing Slurm-based Clusters in Nebius AI We’ve prepared a solution for deploying large-scale, customizable training environments. You get: - Cluster configuration based on NVIDIA stack Quickly deploy a fully prepared cluster with preconfigured NVIDIA CUDA, NCCL, and InfiniBand drivers and libraries. With essential Slurm configurations in place, you can immediately start your high-scale training. - Best training experience If a specific GPU or entire host has issues or underperforms, the Nebius-developed operator mitigates it by providing notifications or recreating the resource. - High-performance shared storage Shared storage, powered by Nebius AI Shared Filesystem, delivers up to 30 GB/s performance for checkpoints and dataset processing. - Adjustable configuration Easily adjust the number of nodes and type of GPUs, and scale your computational resources seamlessly at the start or anytime after deployment. - Easy environment management Users and system packages are synchronized across nodes, simplifying maintenance tasks. - Advanced job scheduling Utilize Slurm for sophisticated job scheduling to optimize resource usage, enhance throughput and minimize job turnaround time. ☝️ By partnering directly with SchedMD LLC, the developer of the Slurm Workload Manager, we provide exceptional support to Slurm users. Explore use cases where Slurm is essential, our documentation, solution library and hands-on videos — then dive straight into the cloud console to create a first VM: https://lnkd.in/d7FcknGY #Slurm #clusters #GPUcloud #orchestrating #scheduling

🧩 Building RAG-based solutions in Nebius AI: all you need to know Retrieval-Augmented Generation offers immense benefits for AI, but implementation can be a challenge. With Nebius Al, you can effectively manage and control the production of RAG solutions, with expert assistance available if needed. Integrate RAG seamlessly into your AI workflows to boost performance and reliability. To those building services using RAG, we offer: - Exceptional UX and a wide range of tools With its intuitive cloud console and tools for Al and RAG workloads such as K8s and Terraform, our platform ensures the best experience. - Marketplace Explore tools from top vendors in machine learning, AI software development and security. Discover the best vector stores and inference tools available. - Best guaranteed uptime Our platform features a self-healing system, allowing VMs and hosts to restart in minutes, not hours. - Scaling your capacity up or down The on-demand payment model allows you to dynamically scale your compute capacity with a simple console request. And you can save on resources with our long-term reserve discounts. We’ve gathered all RAG-related info on one page. Learn more about the architecture, resources and expert support options you can use, and consider our insights during development: https://lnkd.in/eVyHgiSq #RAG #GPUcloud #inference #tools

🌕 JupyterHub: just released on our Marketplace JupyterHub is a multi-user server for Jupyter notebooks, providing a collaborative environment for data science, ML and scientific computing. You can deploy JupyterHub with PyTorch and CUDA in your Managed K8s clusters right away: https://lnkd.in/gyJpmQNs. Just follow the instructions on the page. #JupyterHub #GPUcloud #PyTorch #CUDA

🔥 This Saturday, we’ll be supporting the CUDA Hackathon in San Francisco We’ll provide H100 along with architect support to each hacker participating in the hackathon. Come to build, compete and listen to talks by our speakers: https://lnkd.in/edwtGj7j. It will be great to have Chris Lattner with us, the creator of LLVM, the Clang compiler, and the Swift programming language. Ash Vardanian, Unum’s founder, will also speak at the event, joined by co-hosts Jeremy Nixon, Rohan Pandey and Kyle Morris. If you happen to be in the Bay Area on July 13, we’re hoping to see you! #CUDA #hackathons #events #H100 #GPUs