UNIPATH Universal Communication Data-Path Monitoring and Visualization for Sustainable AI and HPC
- Coordinator
- Koc University, Türkiye
- Runs
- 1 October 2026 to 31 March 2028
- Status
- Running
- EU funding
- €150k
- Total cost
- —
- Partners
- 1 organisations in 1 countries
- Type
- European Research Council grant
- Grant agreement
- 101335111
What the project is doing
In the project's own words, from its summary on CORDIS.
Many of today’s AI and HPC workloads rely on large GPU clusters, and their scalability depends strongly on GPU-to-GPU communication performance. Yet existing tracing tools provide only fragmented visibility into multi-GPU communication. They either capture only one layer of the software stack, expose aggregate network metrics, or remain tied to a single vendor. This makes it difficult to identify communication issues, verify system configurations, and improve performance at scale.
This Proof of Concept project will translate results from our ERC Starting Grant into UNIPATH, a universal, cross-platform communication monitoring and visualization suite for AI and HPC systems. Building on our novel performance tracing tools, UNIPATH will provide end-to-end observability across the communication data path, from application frameworks and communication libraries down to middleware, low-level network operations, and network fabrics.
It will support both CPU- and GPU-initiated communication with low overhead and provide fine-grained visibility into GPU networking activity. UNIPATH will combine AI-assisted performance analysis with interactive visualizations to help system administrators, software engineers, and library developers pinpoint and resolve communication bottlenecks. We will demonstrate its commercial potential through three real-world use cases with early adopters from academia and industry: AI training on over 1,000 GPUs, AI inference on cloud GPU systems, and cross-vendor analysis of a scientific simulation on European supercomputers.
This PoC will position UNIPATH as a distinctive European technology for optimizing communication in sustainable data center infrastructure.
Topics
Who is involved
| Organisation | City | Country | Role |
|---|---|---|---|
| Koc University | Istanbul | Türkiye | Coordinator |
Source: CORDIS, European Commission. Data of 5 Nov 2026, reused under the Commission's reuse policy. View on CORDIS · Grant DOI