
Senior NoC Research Engineer – Network-on-Chip
Posted 5 days ago

Posted 5 days ago
This is a fully remote position, open to applicants in Germany.
• Investigate and assess various NoC topologies, router and switch architectures, as well as arbitration, flow control, QoS, and congestion management.
• Create and uphold architectural models and simulations utilizing SST, gem5, or bespoke cycle-level and analytical frameworks.
• Design traffic generators that accurately represent real-world ML workloads, encompassing collective communication, tensor parallelism, and memory traffic.
• Analyze power, performance, and area (PPA) trade-offs, supporting recommendations with quantitative data.
• Ensure designs are free from deadlock and livelock through thorough analysis, formal methodologies, or simulation techniques.
• Explore adaptive routing approaches and topology alternatives.
• Convert intricate protocol and IP specifications, such as AXI, CHI, and APB, into concise design requirements.
• Evaluate third-party interconnect IP and assist in making build-versus-buy decisions.
• Examine how RDMA principles and chip-to-chip interfaces influence data transfer between accelerators at a system-wide scale.
• Present findings and suggestions to architecture and leadership teams.
• Contribute to the creation of publications and patents focused on on-chip networks and data movement for ML.
• Report directly to the CTO while collaborating closely with chip architecture, hardware, and software teams.
• PhD or a Master's degree with equivalent research experience in Electrical Engineering, Computer Engineering, Computer Science, or a related discipline.
• Proven senior-level experience in NoC or interconnect architecture, either in industry or applied research.
• Solid foundation in on-chip interconnect and traffic management, including NoC topologies, routers and switches, arbitration, flow control, QoS, and congestion control.
• Comprehensive understanding of ML workloads and accelerator architectures, including collective communication, tensor parallelism, and memory traffic.
• Familiarity with on-chip protocols like AXI, CHI, or AHB.
• Knowledge of RDMA concepts and transports (InfiniBand, RoCE), covering queue pairs, memory registration, and zero-copy transfers.
• Experience in evaluating PPA trade-offs using quantitative evidence.
• Proficient in architectural modeling and simulation with SST, gem5, or custom models.
• Strong programming skills in Python and C++.
• Ability to interpret complex protocol and IP specifications and translate them into clear design requirements.
• Experience in deadlock and livelock prevention, adaptive routing, or topology exploration.
• Excellent written and verbal communication skills in English.
• A proactive and collaborative approach to solving open-ended challenges.
• Preferred experience with chip-to-chip and die-to-die interfaces such as PCIe, Ethernet, or UCIe.
• Preferred experience with memory subsystems including HBM and CXL.
• Publications or patents in on-chip networks, interconnect architecture, or data movement for ML or HPC are a plus.
• Familiarity with massively parallel architectures is advantageous.
• Applicants are required to provide their current location and indicate whether visa sponsorship and/or relocation assistance is necessary to work in Germany.
• 30 days of annual vacation.
• Up to 2 weeks of work abroad.
• Engage in cutting-edge sparse computing technology.
• High-impact position with a direct influence on the architecture of SpiNNext.
• Collaborative team environment with seasoned engineers from academia and industry.
• Opportunity for equity participation in a growing deep-tech enterprise.
Dragos, Inc.
Dragos, Inc.
Ramp
Get handpicked remote jobs straight to your inbox weekly.