Explainable Network Topology Recommender for the HPC Cluster Configurator

Brief: Build an explainable recommender inside the HPC Cluster Configurator focused on Fat-Tree/leaf-spine network topologies for HPC/GPU clusters.

Goals and tasks:

• Model network topologies; design input schema (nodes, NICs per node, link speed, switch ports, target oversubscription).
• Compute port usage, uplink/downlink bandwidth, scalability, oversubscription.
• Validate configurations; rank suitable options; generate clear, user-facing explanations.
• Prepare test sets and document formulas, assumptions, and limitations.

Profile/skills:

• Strong Python; knowledge of AI/LLM prompting or explainable recommendation; basic networks knowledge; interest in HPC; good documentation skills.
• Nice-to-have: Fat-Tree/Clos, InfiniBand/Ethernet/RDMA, REST APIs, basic frontend integration.

Duration & perks:

• Target: students entering final year.
• Period: 2–3 months.
• Paid: Monthly stipend.