Software Engineer (Scale Out)
Posted Updated
- In this role, you’ll contribute to the development and optimization of the infrastructure that powers our cutting-edge processors, with a primary focus on C/C++ development and low-level programming
- You’ll work closely with large inference and training model development to further drive Scale Out software and hardware performance
- Design, develop, and maintain core infrastructure software that interfaces directly with Tenstorrent hardware
- Build low-level libraries and APIs for communication and synchronization across compute nodes
- Optimize system-level software for performance, scalability, and reliability in distributed environments
- Support hardware bring-up, system testing, and deployment in close collaboration with hardware and software teams
- Integrate infrastructure components with internal programming models, tools, and runtime systems
- What You Will Learn:
- The execution model and system architecture of Tenstorrent AI hardware
- How distributed accelerator systems coordinate communication and synchronization at scale
- Advanced techniques in system-level performance tuning across multiple layers of the stack
- How low-level infrastructure, runtimes, and hardware co-evolve in high-performance AI platforms
- The challenges of building reliable, production-grade infrastructure for large-scale AI workloads
- Tenstorrent is seeking a skilled Software Engineer with a passion for building high-performance, low-level systems software
- This role is hybrid, based out of Toronto, ON
- Comfortable working close to hardware and curious about how systems behave under the hood
- Experienced building low-level software, drivers, embedded systems, or performance-critical infrastructure
- Structured problem solver who thrives in fast-paced, highly technical environments
- Strong C or C++ systems engineer with a deep understanding of memory, threading, I/O, and low-level execution models
- Proficient with Linux systems programming and debugging tools such as gdb, strace, and perf