About the Role
As a Runtime Engineer, you will be responsible for the execution layer of our platform, ensuring the code generated by our compiler runs with maximum efficiency and accuracy on varied hardware architectures. This role demands deep, low-level systems expertise covering thread scheduling, parallel workload distribution, hardware benchmarking, and cross-functional engineering partnerships.
Key Responsibilities:
- Architecture Development: Architect, implement, and support our multi-platform execution engine.
- Workload Optimization: Utilize advanced partitioning and concurrent execution strategies to automate instruction scheduling and leverage high-performance hardware paths.
- Profiling & Analysis: Measure and evaluate compiled binaries directly on physical hardware to guarantee top performance.
- Diagnostics: Create internal tools and telemetry to pinpoint and resolve execution stalls and hardware bottlenecks.
- R&D Exploration: Build functional prototypes and gather data to test experimental execution mechanisms.
- Product Alignment: Interface with product managers to convert high-level machine learning workloads into concrete systems-level enhancements.
Key Requirements:
- Education: Degree in Computer Engineering, Computer Science, or an equivalent technical field.
- Domain Knowledge: Professional experience developing runtime systems, execution engines, or hardware-facing compiler backends.
- Programming Skills: Exceptional proficiency in modern systems-level programming (C++14 and above).
- Concurrency: Thorough mastery of multi-threaded, asynchronous, and concurrent software architectures.
- Hardware Insight: Strong grasp of physical computer architecture, including memory caching tiers, SIMD/MIMD instructions, and processor registers.
- Operating Systems: Practical familiarity with operating system primitives, driver layers, or low-level kernel mechanics.
- Bonus Points: Hands-on experience with GPGPU frameworks (CUDA/ROCm), machine learning compilers (Triton, PyTorch, JAX), or distributed High-Performance Computing (HPC) clusters.
