A complete application in a minute
Tailored resume and cover letter, ready to send.
Callosum is seeking a Hardware-Aware Model Optimisation professional to reshape large language models for diverse hardware accelerators. The role emphasizes building efficient inference workflows and rigorous experimentation to ensure real-world improvements in quality, speed, and cost.
You will work at the intersection of AI models and silicon, exploring new chips and architectures, and developing automated methods to test and validate optimized variants for deployment.
The role asks for someone who can reshape large language models so they run well on many different kinds of chips, not just standard GPUs. A strong candidate brings deep knowledge of how modern models are built and run, plus hands on work in efficient inference or hardware aware machine learning. They also need a habit of designing careful experiments that show whether a modified model is actually better on quality, speed and cost.
The work sits at the point where models and silicon meet, and much of it is about adapting models to new chips rather than the other way around. AI can help search and test many model variants quickly, but judging which trade offs are real and worth shipping stays with people. The human part is framing the right experiments and reading results with care.
A person here would gain direct practice in tuning models for a wide range of accelerators, including the named chips the company works with, and in building automated ways to find and check specialised model variants. That experience comes from running real experiments on heterogeneous hardware and measuring quality and cost side by side. It is a strong base for work on efficient inference and model architecture research.
This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.
Callosum builds infrastructure that routes artificial intelligence workloads to the specialized model and silicon chip most suitable for each task, per the company.