The posting
Description des tâches:About us:LuxProvide is pioneering the future of digital intelligence, merging data for science with high-performance computing for accuracy, security, and sustainability. At the center of our work is MeluXina supercomputer, soon to be joined by MeluXina-AI and MeluXina-Q (quantum), expanding our capabilities into next-generation AI and quantum computing. As a key partner in the Luxembourg AI Factory, we play a central role in enabling researchers, startups, and industry to develop and deploy trusted AI solutions at scale. Strategically based in Europe's digital hub, LuxProvide connects research and industry and drives innovation both within Luxembourg and across borders. Join us and be part of a fast-evolving ecosystem where supercomputing, AI and quantum converge to shape breakthroughs that will define tomorrow.Senior AI Software Engineer (F, M, X)The AI Software Engineer will join the expert teams managing flagship platforms at the heart of Luxembourg's and Europe's supercomputing, AI and quantum strategies – MeluXina, MeluXina-AI and MeluXina-Q. This is a unique opportunity to shape the AI capabilities of one of Europe's most advanced supercomputing and AI environments.As part of the Supercomputing Application Services team, you will own the performance, quality and cost-efficiency of our AI-based applications, including GenAI/LLM, acting as the organization's go-to expert on AI, including AI models and LLMs in particular. You will ensure our AI platforms run at the frontier of what is technically and economically achievable on our GPU-accelerated infrastructure, on MeluXina and the upcoming MeluXina-AI platform.You will perform deep-tech work, driving optimization of AI inference and development, including model training, fine-tuning, benchmarking and performance engineering, together with championing best practices for customers and across the organization.What your role will consist inResponsibilitiesAI excellence• Optimize the design and deployment of GPU-accelerated AI model inference workloads, maximizing the performance-to-cost ratio for the most demanding AI projects.• Own the deployment and operation of inference engines, with an initial focus on vLLM, TensorRT-LLM and cutting-edge tooling such as NVIDIA Dynamo, Triton and NIMs.• Design and perform LLM fine-tuning to meet the specific needs of MeluXina-AI customers, with attention to performance and model quality.• Design and implement evaluation and benchmarking frameworks for AI systems, assessing model quality, performance, reliability, safety and business value.• Design, develop, test and maintain software components, internal tools, automation and reference environments supporting AI workflows, model serving and benchmarking.• Advise teams and customers on AI architecture, model selection and use cases, while designing and implementing AI-powered solutions including RAG, agents, multimodal systems and enterprise integrations.• Lead user enablement activities by providing expert guidance, troubleshooting, and optimization advice for the Mistral AI platform.• Establish and drive best practices for model evaluation, serving and responsible AI, while implementing MLOps/LLMOps capabilities including versioning, experimentation, monitoring and lifecycle management.• Ensure AI applications remain reliable, scalable, secure and compliant through continuous monitoring, optimization and governance practices.• Keep the organization at the forefront of LLM and AI research, translatingadvances into practical improvements shared across the organization and to our customers.Software stacks:• Deploy, maintain and evolve AI software stacks supporting the Company's supercomputing, AI, quantum and cloud platforms, balancing common services with customer-specific requirements.• Drive the adoption of Cloud-native functionalities of the Company's platforms, including Kubernetes, OpenShift, and related ecosystems, to benefit internal and customer applications.• Contribute to software quality through code reviews, CI/CD practices, functionality testing and performance validation of AI, HPC and QC software.• Develop internal and user-oriented tools to improve the management of software environments and the user experience.Customer engagement, projects and collaboration:• Provide advanced support and solutions to Company's customers, engage with customers to answer complex questions that come from using or porting workflows, pipelines and applications to the Company AI and Cloud platforms.• Evaluate and support national, European and international projects coming to the Company's platforms, to ensure optimal use of the facilities for excellence in science and innovation.• Collaborate with solution providers and organizations having a similar or complementary mission to the Company, to enhance the delivery of services to Company's customers and develop new products and services.ProfileQualifications & Experience• Advanced degree (Master or PhD) in Computer Science, Machine Learning, Data Science, Engineering, or a related field.• Strong experience working with large language models and modern AI frameworks in production or large-scale environments.• Solid theoretical understanding of LLMs and AI, with the ability to reason about architecture and trade-offs.• Practical experience with software deployment, performance profiling and troubleshooting.Professional ExpertiseMust-have:• 5+ years of experience in AI/ML engineering, MLOps, software engineering or related domains.• Proven experience running large-scale distributed training, fine-tuning and inference workflows on multi-GPU and multi-node infrastructures, including: o supporting reliable model serving; expertise with vLLM or comparable inference engines in production is a must. o API-based deployment, autoscaling, load balancing, monitoring, versioning, and the secure exposure of endpoints.• Experience optimizing AI workloads on NVIDIA GPU platforms, including profiling, quantization, parallelism techniques and accelerator-aware deployment.• Experience supporting AI workloads across the full lifecycle, from experimentation and data preparation to training, fine-tuning, evaluation, deployment, monitoring, and continuous improvement.• Deep understanding of modern AI concepts including model training, fine-tuning (FT), retrieval-augmented generation (RAG), reinforcement learning (RL), model evaluation and inference optimization.• Knowledge of containerized and cloud-native environments, including Kubernetes / OpenShift-based services, orchestration, image management, observability, and secure multi-tenant deployment patterns.• Proven software development skills in Python.Great-to-have• Experience with model output validation, consistency, and formalization.• Experience with Supercomputing/HPC environments, HPC and AI software stacks and middleware, including schedulers (Slurm/Flux), distributed applications, parallel filesystems and large-scale GPU infrastructure.• Great proficiency with C++, CUDA or other performance-oriented languages.• Experience preparing, curating and evaluating datasets for AI model training and assessment.Communication & Collaboration• Ability to communicate complex LLM and AI concepts clearly to technical and non-technical audiences (co-workers and users).• Strong problem-solving skills, with a pragmatic and solution-oriented approach.• Excellent communication and collaboration skills in international environments.• Demonstrated success advising cross-functional teams and promoting LLM best practices across organizations.Personal Attributes• Curious and rigorous, with a strong interest for LLMs and AI.• Demonstrated interpersonal skills, fostering positive team dynamics and respondingconstructively to challenges.• Strong organizational [...]
