Altos is an integrated hardware and software AI computing platform that streamlines development environments and hardware management. It pairs computing systems such as Altos BrainSphere workstations and servers with the Altos Accelerator Resource Manager to coordinate CPU, GPU, memory, and storage allocation. The system automates resource discovery and schedules computing tasks while deploying necessary AI software stacks and dependencies automatically. With support for NVIDIA Multi-Instance GPU technology, Altos allows multiple users to partition high-performance cards into isolated instances, preventing configuration conflicts and hardware idling across shared research or engineering environments. It also incorporates GPU topology-aware scheduling to place data-heavy workloads along optimal communication paths, reducing transfer bottlenecks between processors and system memory. Team members can access hardware clusters and track workloads remotely via a web-based management console from any standard browser. Altos is intended for AI developers, data scientists, and IT infrastructure managers who need to optimize hardware usage and manage shared compute infrastructure. Pricing information for Altos is not publicly provided.
Problem: In university research labs or small AI startups, multiple researchers often need to share a limited number of high-performance GPUs. Manually scheduling "who uses which card" leads to hardware idling, resource conflicts, and the tedious task of manually reinstalling different software versions (like different CUDA or PyTorch versions) for each project.
Solution: Altos aiWorks uses the Altos Accelerator Resource Manager (AARM) to create independent development spaces for each user. It supports NVIDIA A100 Multi-Instance GPU (MIG) technology, allowing a single powerful GPU to be partitioned into smaller, isolated instances.
Example: A lab manager configures an Altos BrainSphere R680 server. One student requests a slice of the GPU for a small NLP model, while another requests a larger portion for computer vision training. AARM automatically schedules these resources and deploys the specific software stacks each student needs without them interfering with each other.
Problem: AI developers spend significant time configuring drivers, libraries, and dependencies every time they start a new project or switch from a training to an inference environment. This "environment hell" slows down the development cycle and increases the barrier to entry for new team members.
Solution: The solution provides a "One-Stop" hardware and software integration that allows for the rapid deployment of development environments. AARM automates the discovery and scheduling of CPU, GPU, memory, and storage, deploying the necessary AI software stacks in the backend.
Example: A startup developer needs to quickly test a new LLM. Instead of manually setting up the server, they use the AARM web console to select a pre-configured template. The system automatically provisions an Altos BrainSphere P550 workstation with the required environment, allowing the developer to start coding in minutes rather than hours.
Problem: Data scientists often require the power of a high-end workstation or server (like the BrainSphere R385), but they cannot always be physically present in the office or data center. Standard remote desktop solutions can be laggy or difficult to manage for complex AI workloads.
Solution: Altos aiWorks features a Web-Based Management Console that supports remote login. This lifts the restriction of local hardware resources, allowing developers to access high-performance computing power from any location using a standard web browser.
Example: A lead data scientist working from home logs into the Altos web console to monitor a long-running training job on an Altos BrainSphere R685 F5 server located at the company headquarters. They can adjust resource allocations or redeploy a new environment remotely without needing physical access to the server room.
Problem: Growing businesses often over-invest in hardware that is underutilized because they lack the tools to manage "GPU topology-awareness." Without smart management, data transfer bottlenecks between CPUs and GPUs can cause significant overhead, wasting the expensive hardware's potential.
Solution: Altos aiWorks provides GPU topology-aware resource allocation. By understanding how the hardware is physically interconnected, AARM minimizes communication overhead and ensures that tasks are assigned to the most efficient resource paths.
Example: An enterprise is running multiple concurrent AI tasks across a cluster of BrainSphere R389 F4 servers. AARM analyzes the GPU topology to place data-heavy tasks on GPUs that have the fastest direct paths to the necessary memory and CPU cores, maximizing the Return on Investment (ROI) by getting more processing power out of the same hardware.
Target audience: Best for: AI developers, IT infrastructure managers, Data scientists
Pricing: Unknown · Categories: Startup tools
Tags: startup tools
Altos is an integrated hardware and software solution that combines systems like BrainSphere servers with the Altos Accelerator Resource Manager. It automates hardware discovery, allocates CPU, GPU, memory, and storage, and automatically deploys software stacks so teams can build, test, and run machine learning workloads without manual configuration bottlenecks.
Altos is designed for artificial intelligence developers, data scientists, and IT infrastructure managers. It supports academic research laboratories, startups, and enterprise engineering departments that must share high-performance workstations and server clusters among multiple users who require distinct software dependencies and isolated development environments.
Altos uses GPU topology-aware resource allocation alongside NVIDIA Multi-Instance GPU support. The Altos Accelerator Resource Manager identifies the physical interconnects between processors and system components, scheduling tasks across the fastest data paths. It can also divide compatible GPUs into independent, isolated instances to let multiple developers work concurrently without interference.
Altos features a web-based remote management console accessible via standard web browsers. Distributed teams can log into centralized BrainSphere hardware clusters remotely to provision pre-configured software environments, submit training tasks, adjust compute allocations, and monitor long-running workloads without needing direct physical access to the server room.
Altos resolves software conflicts by providing multi-user independent development environments. When a user requests computing resources, the Altos Accelerator Resource Manager automatically deploys the requested software stack and libraries in an isolated space, allowing team members to run differing framework or CUDA versions on the same shared cluster simultaneously.