Solution Architect - Kernels and Performance, Core ML

We're looking for a Solution Architect – Kernels and Performance, Core ML to join our team in Lisbon, Portugal, in a hybrid working mode. In this role, you will drive low\-level performance engineering of AI workloads, optimizing both model training and inference across advanced accelerator architectures like TPU and GPU. You will work on cutting\-edge ML models, toolchains and frameworks, enabling scalable, efficient deployment of AI solutions in production. This position combines deep system\-level engineering with architectural leadership, directly impacting next\-generation AI performance.

Responsibilities

• Design and optimize high\-performance kernels using low\-level languages like Pallas, Mosaic and Triton for TPU and GPU architectures


• Architect infrastructure such as benchmarking suites, autotuning frameworks and performance analysis tools to support kernel development and testing


• Develop regression testing strategies and comprehensive documentation to maintain quality and facilitate adoption across developer communities


• Collaborate with ML researchers, framework developers (JAX, PyTorch) and compiler engineers (XLA) to address performance bottlenecks and implement effective solutions


• Track advancements in hardware architectures, compiler technologies and AI models to identify optimization opportunities and guide roadmap decisions


• Advocate best practices for integrating optimized kernels into open\-source libraries and production systems

• Bachelor’s degree in Computer Science or equivalent practical experience


• 12\+ years of overall industry experience in software engineering or related fields


• Minimum 5 years of experience in C\+\+ or Python development


• At least 3 years of experience testing, maintaining or launching software products


• Minimum 1 year of experience in software design and architecture


• Proven background in kernel\-level performance optimization for ML workloads

• Experience optimizing TPU/GPU code using Pallas, CUDA or Triton


• Familiarity with ML frameworks such as JAX and PyTorch, including advanced components like attention mechanisms and Mixture of Experts (MoEs)


• Understanding of modern accelerator characteristics such as data movement, pipelining and heterogeneous compute


• Knowledge of compiler principles, code generation and toolchains such as MLIR and OpenXLA


• Experience building developer infrastructure for OSS libraries and creating high\-performance APIs


• Strong problem\-solving and investigative skills with proven ability to work across cross\-functional teams

• Competitive compensation depending on experience and skills


• Variety of projects within one company


• Being a part of a project following engineering excellence standards


• Individual career path and professional growth opportunities


• Internal events and communities


• Flexible work hours

Este anúncio é de indeed. Ver anúncio original ↗