Storage that Keeps Up with
Your Research
In higher education research environments from genomics, to climate modeling, to physical simulation, NeuralMesh™ delivers a high-performance storage solution built for modern university research clusters. Collapse infrastructure silos, reclaim idle GPUs, and move from data ingestion to discovery without waiting on storage.
Leading research universities are modernizing storage infrastructure with NeuralMesh.
Faster Research, Lower Cost from the GPUs You Own
Legacy parallel file systems were built for sequential I/O, but modern AI workloads require storage that can support metadata-heavy, small-file, IOPS-intensive patterns that define today’s research. As HPC and AI converge on shared clusters, your storage becomes the bottleneck. NeuralMesh™, NeuralMesh™ Axon™, and Augmented Memory Grid™ deliver the throughput and metadata performance to keep your GPUs fed and your research moving.
Collapse Silos into
Shared Capacity
Converge isolated systems into one shared resource that reclaims stranded GPUs and enforces tenant isolation
Serve Every Workload from One Namespace
Eliminate data copies by storing small and large files, training and archive data on NeuralMesh across POSIX, NFS, SMB, and S3
Tier Hot and Cold
Data Automatically
Layer fast flash over cheaper disk-based capacity tier, promote cold data automatically, and keep any archived dataset ready
on demand
Scale from Terabytes
to Exabytes
Grow predictably without re-architecting as research intensifies, holding metadata performance steady under billions of files
Maximize GPU Utilization
Push utilization from 30% on legacy systems to more than 90% by keeping every application node continuously fed
Accelerate Checkpointing
Absorb failures and remove job stalls enabling long training runs to complete without losing GPU hours to restarts
Shrink the Data
Center Footprint
Cut racks, power, and cost per watt while increasing performance density
Federate and Share
Data Securely
Move data fast across institutions and dispersed teams while maintaining performance and supporting
inter-institution research
Simplify Operations for
Lean Teams
Replace manual tuning and staging with policy-driven management so staff spend less time on maintenance and more time
supporting research
NeuralMesh Delivers Speed and Scale to Accelerate Every Research Domain in Higher Education
AI training, genomics, climate modeling, and physical simulation all demand high-throughput storage with low-latency metadata. NeuralMesh serves every workload from a single namespace without tuning or without compromise.
Scientific Simulation and Mmodeling
Climate, CFD, astrophysics, materials, and quantum simulation mix large sequential reads with bursty parallel I/O across many nodes. NeuralMesh serves both patterns without the frequent tuning trade-offs.
Genomics, Bioinformatics, and Imaging
Petabyte-scale sequence datasets with billions of small files make metadata the bottleneck. NeuralMesh handles it natively and moves imaging data directly into GPU pipelines without extra copy steps.
Physical AI, Robotics, and Computer Vision
Synthetic data generation with Isaac and Omniverse, video analysis at scale, and small-file labeling pipelines all run on the same system without requiring a separate storage tier.
Shared Research Data Services
Multi-tenant institutional clusters serving many departments need fair-share, chargeback support, and fast federation across institutions, hybrid, and cloud. NeuralMesh delivers all three.
AI and Foundation Model Training
Campus AI factories train and fine-tune models across many concurrent groups. NeuralMesh supports long training runs with rapid checkpointing and continuous read/write for agentic and retrieval workloads.
NeuralMesh Offers a Proven Solution for Higher Education Research
“With NeuralMesh, we can expand input/output capabilities to ensure that even the most data-intensive research projects are supported efficiently.”
“WEKA came forward and promised to do these amazing IOPS and so far so good. It was just smooth. There were no complaints about the storage, it all just worked. And that to us is a big win.”
“With NeuralMesh, our researchers no longer have to worry about what’s happening in the background when they’re running their experiments.”
“HPC tends to turn CPU problems into I/O problems. NeuralMesh turned those I/O-bound problems back into CPU-bound problems.”
“We immediately saw our cloud storage costs drop by 90% when we switched to WEKA.”
“We’ve actually got a solution which is going to work for the next three or four years at least as we grow. And that’s fantastic.”
“Implementing NeuralMesh has made our jobs easier. Previously, a lot of time was spent managing the cluster’s file system, but now it’s almost set-and-forget.”
“Going from the old NFS solution to NeuralMesh has been such an upgrade. The amount of throughput we get allows researchers to meet their publication deadlines.”
“We can support 6x the amount of research projects and are still growing. WEKA has unlocked a lot of research potential for us.”
“It frees up the team to focus on other problems within the cluster and other things that are going on. To have that freedom during that deadline season was really key for us.”
“Implementing NeuralMesh has given the institute the capability to keep pace with the growing demands of our researchers.”
Choose NeuralMesh to Accelerate Research in Higher Education
When legacy storage stalls your research, NeuralMesh unifies your data and frees your teams to move faster. Ready to see what’s possible with NeuralMesh?
Common Questions, Straight Answers
NeuralMesh™ by WEKA is purpose-built for university HPC and AI research. It delivers high-throughput, low-latency storage across POSIX, NFS, SMB, S3, and GPUDirect Storage from a single namespace running genomics, AI training, and simulation on shared clusters without per-workload tuning. Learn more at weka.io/solutions/higher-education-research.
Legacy parallel file systems starve GPUs with slow metadata and small-file I/O. NeuralMesh eliminates those bottlenecks, keeping every GPU node fed. Deakin University saw a 10X performance improvement, with training epochs dropping from 40–60 minutes to roughly 6 minutes. See WEKA’s research solution brief.
Yes. NeuralMesh runs HPC simulations and AI training from a unified namespace, supporting GPUDirect Storage for AI pipelines alongside large sequential reads for climate modeling, CFD, and genomics. No configuration trade-offs required. Learn more at weka.io/product/neuralmesh.
NeuralMesh delivers full tenant isolation with per-filesystem access policies, supporting fair-share and chargeback across dozens of concurrent research groups with no cross-tenant performance degradation. The Wharton School runs 20+ research centers on a single NeuralMesh cluster in the cloud. More at weka.io/customers/the-wharton-school.
Genomics pipelines generate petabyte-scale datasets where metadata performance is the constraint. NeuralMesh strips metadata across all cluster nodes for consistent performance at scale. WEKA’s platform has allowed labs to achieve 3x faster genomic workloads vs. GPFS.
NeuralMesh automatically tiers data between NVMe flash and object storage. Cold datasets archive and promote instantly when a project resumes without manual restaging. A single cluster scales to 14 exabytes with up to 1,024 filesystems, each with its own tiering policy. Learn more at weka.io/product/neuralmesh.
NeuralMesh is a software-defined parallel file system that replaces GPFS and Lustre with faster metadata, native small-file performance, and full protocol support without legacy tuning complexity, and the same software runs in all the major public clouds. In one customer use case, Wharton replaced legacy NFS on AWS and gained throughput to every compute node.
NeuralMesh replaces manual staging and cluster tuning with policy-driven, automated management. After deploying NeuralMesh, Deakin University’s AI Systems Administrator called it “almost set-and-forget,” freeing lean research computing staff to support researchers instead of maintaining infrastructure. More at weka.io/customers/deakin-university.