Software AI Model Optimisation & Benchmarking Engineer

About the Role

We are seeking aMachine Learning / AI Engineerwith expertise inmodel optimisation, acceleration, and deployment on AI accelerators(GPUs, NPUs, ONNX-compatible hardware). You will work onalgorithm development, performance tuning, and benchmarkingof a new AI accelerator to different AI applications (LLM's/ViT's/DiT's/CompVision). Knowledge ofoptical computing and photonics-based AI accelerationis highly beneficial.

Key Responsibilities

• Develop, optimise, and accelerate ML models usingTensorFlow, PyTorch, and ONNX.
• Implementquantisation, pruning, and other optimisation techniquesto improve model efficiency.
• Benchmark AI models on various accelerators (GPUs, NPUs, TPUs, optical computing platforms) and fine-tune performance.
• Optimise inference and training pipelines for speed, energy efficiency, and hardware compatibility.
• Work withCUDA, cuDNN, TensorRT, and other low-level libraries for AI acceleration.
• Explore and applyoptical computing techniquesfor next-generation AI acceleration.
• Collaborate with software, hardware, and optical computing teams to ensure seamless deployment.

Requirements

• Strong experience inPythonwithTensorFlow/PyTorch.
• Experience inAI model optimisation techniques(quantisation, pruning, knowledge distillation).
• Knowledge ofAI accelerators(GPUs, NPUs, ONNX, TensorRT, OpenVINO).
• Hands-on experience withbenchmarking toolsand performance profiling.
• Understanding of parallel computing, memory optimisation, and hardware-aware ML development.
• Familiarity withoptical computing concepts(photonics, optical neural networks) is a plus.

Nice to Have

• Experience withlow-level CUDA programmingfor deep learning.
• Familiarity withedge AIand deploying models on embedded systems.
• Exposure tooptical-based AI accelerationand photonics-driven hardware.
• Experience with large-scale distributed training (e.g., Horovod, DeepSpeed).

Contact Jonny for more information.