Job Summary:
Etched is building AI chips that are hard-coded for individual model architectures, and they are seeking a Software Engineer to help optimize and implement software for their LLM compilation. The role involves writing optimized kernels for transformer operations and integrating with existing libraries to ensure compatibility with their chip.
Responsibilities:
โข Write an optimized kernel to compute a new attention variant on our hardware
โข Implement HuggingFaceโs `CohereForCausalLM` class using Etchedโs transformer building blocks
โข Implement a synchronization mechanism to coordinate between the host CPU and Etched accelerator
โข Implement FP8 quantization for FP16 models using the same mechanism as TransformerEngine
Qualifications:
Required:
โข Have 3+ years of software engineering experience
โข Have experience working with machine learning operators
โข Are comfortable doing low-level embedded programming
โข Pick up slack, even if it goes outside your job description
โข Are results-oriented, and bias towards shipping products
โข Want to learn more about machine learning research
Preferred:
โข Transformer optimizations, such as FlashAttention
โข Ongoing research in machine learning
Company:
**Acquired by OneCruit in July 2025** OpenReq is the embedded recruiting firm built for early-stage startups (Seed to Series B) in the AI & Hard Tech space. Founded in 2020, the company is headquartered in San Diego, USA, with a team of 11-50 employees. The company is currently Early Stage.