דרושים » תוכנה » מוביל /ת הטמעה ופיתוח כלי AI (פקולטי) בפקולטה להנדסת חשמל ומחשבים, JB-2542

Our team at the Huawei Computing Network Innovation Lab is looking for exceptional talent to join us and lead the development of next-generation data centers. We create cutting-edge technologies that synergize software and hardware in tandem to accelerate compute, storage, and networking at large scale. We aim to drive innovation and deliver software-defined infrastructure and algorithms for HPC, AI/ML, and Big Data applications.
We are looking for an outstanding Research Team Lead with deep hands-on expertise in large-scale distributed systems, AI framework infrastructure, and performance optimization on custom accelerators. If you are a visionary technical leader, a skilled communicator, and a team builder who thrives at the frontier of systems and AI research - you're welcome on board.

What Will You Be Doing?
Lead and grow a world-class team of researchers and engineers working on distributed AI infrastructure and systems software.
Architect and own the software infrastructure enabling distributed training and inference on Huawei's custom accelerator hardware (e.g., Ascend NPU).
Drive research and development on communication libraries, runtime systems, memory management, and graph execution & synchronization at scale.
Optimize end-to-end performance across large-scale clusters, covering both scale-up (multi-device) and scale-out (multi-node) configurations.
Design and implement high-performance communication backends and collective operations (AllReduce, AllGather, broadcast) for distributed training workloads.
Collaborate cross-functionally with hardware architects, compiler teams, and framework engineers to co-design hardware-software solutions.
Publish and present research findings at top international venues (NeurIPS, EuroSys, SC, MLSys, OSDI) and represent the team externally.
Mentor engineers and researchers, conduct performance and growth reviews, and shape team culture and technical direction.
Partner with top academic institutions and open-source communities to advance the state of the art in distributed AI systems.

Are you an inventive, curious, and driven Applied Scientist with a strong background in AI, Computer Vision, and Deep Learning? Join our AGI IMAX Science team and contribute to significant advancements in Computer Vision, Multimodal Understanding, Generative AI, and foundational models.

As part of the AGI IMAX Science team, you'll lead innovative research projects and train large-scale Vision-Language Models (VLMs), diffusion models, and multimodal foundation models that directly impact millions of Amazon and AWS customers. Leveraging Amazon's vast computing power, you'll work alongside a supportive and diverse group of skilled scientists and engineers, building models and services that make a meaningful difference in the industry.


Key job responsibilities
Lead research initiatives in Computer Vision and Multimodal generative AI, advancing model efficiency, accuracy, and scalability.
Train and fine-tune large-scale Vision-Language Models (VLMs), diffusion models, and multimodal foundation models at scale.
Design, implement, and evaluate deep learning models in a production environment.
Collaborate with cross-functional teams to transfer research outcomes into scalable AWS services.
Publish in top-tier conferences and journals, keeping Amazon at the forefront of innovation.
Mentor and guide other scientists and engineers, fostering a culture of scientific curiosity and excellence.

Are you passionate about systems programming at the intersection of hardware and software? Do you want your code to directly accelerate the world's largest AI training and inference workloads? The Elastic Fabric Adapter (EFA) Drivers team is looking for Software Development Engineers to design and implement the kernel drivers and userspace libraries that power high-performance networking for AWS's AI/ML and HPC infrastructure.

EFA is the custom network interface that enables thousands of GPUs and accelerators to communicate at near-wire-speed bandwidth and low latency, making it a critical enabler of both foundation model training and real-time inference serving at unprecedented scale. When customers train the next generation of large language models or serve billions of inference requests with tight latency SLAs on our EC2 P5, P6, or Trn instances, it's our driver stack that moves the data.

Key job responsibilities
- Design, develop, and optimize Linux kernel drivers (RDMA/EFA) and userspace provider libraries (rdma-core) that ship to millions of EC2 instances.
- Work directly with custom hardware - collaborate with chip designers to bring new silicon capabilities to life in software.
- Contribute to the upstream Linux kernel RDMA subsystem and rdma-core open-source project.
- Architect solutions for next-generation networking features: GPU-direct RDMA, adaptive routing, collective offloads, and multi-path transport.
- Build monitoring and automation tools to enhance driver testing for functionality, reliability, and performance.
- Own the full lifecycle: from initial hardware bring-up and feature enablement through testing, deployment, and production troubleshooting across the global AWS fleet.

Looking for a CPU performance modeling architect to take responsibility over the performance aspects of new CPU instructions or modes of operation. The role includes but is not limited to:
Partners with lead SW and HW architects to co-invent optimal HW and SW solutions that come to address requirements. Influences the direction based on experiments and simulation data.
Models CPU functionality, performance and/or power in pre-silicon simulators
Defines and runs performance experiments to aid feature definition. Such experiments can be performed on a pre-silicon simulation environment or in a real system or on a combination of both or even in combination with analytical models.
Provides experimental/proof of concept for new features and implementation alternatives meeting performance constraints.
Analyzes the bottlenecks of current CPUs on workloads that reflect CPU future usages.
Potentially (in the future), lead a team doing above activities.
An adequately qualified candidate can also become the leader of definition for some features in addition to all the roles above.