TT-Fabric Software Engineer
Posted Updated
- Tenstorrent is building the world’s fastest, most efficient AI compute clusters. TT-Fabric is the high-performance nervous system of this platform: the low-level networking layer that lets thousands of RISC-V and AI processors snap together into a single, massively parallel distributed supercomputer
- Architect, implement, and maintain TT-Fabric, our low-level networking library powering distributed inference and training
- Design scalable communication systems capable of coordinating thousands of AI processors efficiently and reliably
- Optimize protocols, synchronization strategies, and data movement to extract maximum hardware performance
- Integrate TT-Fabric APIs into the broader programming model in collaboration with AI and hardware teams
- Help define the long-term architecture of Tenstorrent’s distributed systems stack
- What You Will Learn
- How large-scale AI clusters are architected from the networking layer up
- The performance characteristics of custom AI hardware and RISC-V processors at scale
- Advanced synchronization, collective communication, and interconnect optimization techniques
- How distributed systems design decisions directly influence model throughput and training efficiency
- How hardware and networking software co-evolve in next-generation AI infrastructure
- We welcome candidates at various experience levels for this role
- Interest in networking protocols, synchronization and communication in large clusters of computers, and high-performance interconnects
- Bachelor’s degree or higher in Computer Science, Electrical/Computer Engineering, or a related field
- Willingness to think from first principles, consider out-of-the-box solutions, discover where industry norms and current state-of-the-art fall short, and drive to surpass them
- Passion for the intricacies of low-level programming and optimizing hardware-software interactions at the bare-metal level, where every nanosecond counts
- Strong foundation in C/C++, systems programming, or low-level bare-metal programming
- Experience with programming large-scale computer clusters is a plus but not required
- Experience with high-performance networking libraries, RDMA, or MPI is a plus but not required
- During the interview process, candidates will be assessed for the appropriate level, and offers will align with that level, which may differ from the one in this posting