AI Research Engineer (Kernel & Inference Optimization) - 100% Remote Worldwide (Tether)

Optimize model serving pipelines and inference frameworks for high-throughput, low-latency, low-memory-footprint AI performance. Establish performance metrics and identify production bottlenecks. Develop custom compute shaders and GPU kernels for smartphones. Build and deploy end-to-end inference pipelines on resource-constrained devices. Apply empirical research to overcome model serving challenges across distributed inference systems.

Computer Science degree or related background, ideally a PhD in NLP or Machine Learning with strong A-tier conference publications. Proficiency in Metal Shading Language with the ability to write custom compute shaders. Demonstrated low-level kernel optimization on mobile devices with measurable latency, throughput and memory gains. Deep knowledge of modern model serving architectures and inference optimization. Understanding of distributed inference using Tensor, Pipeline and Expert Parallelism. Mathematical understanding of Diffusion Models and Vision Transformers. Knowledge of pruning, quantization, flash attention, KV cache and speculative decoding.