We are seeking a highly skilled and experienced AI IP Performance Engineer to join IP Product team. This team is focused on managing Accelerated Compute IP products with expertise in RISC-V, and AI IPs. As an AI IP Performance Engineer in this team, you will work closely with the AI architecture team in India to prototype ideas, analyze performance, energy, and cost trade-offs across hardware/software options. You should have a firm grasp of deep learning workloads, especially the large generative LLM/Diffusion models.
This role is hybrid, based out of Bengaluru, IN.
We welcome candidates at various experience levels for this role. During the interview process, candidates will be assessed for the appropriate level, and offers will align with that level, which may differ from the one in this posting.
Responsibilities:
Lead the performance power measurement and benchmarking activities
Develop performance power models for competitive benchmarking and roadmap projections
Collaborate with cross-functional teams to drive architectural innovations from model insights to implementation.
Explore HW/SW Optimizations at Component, SoC, Network (Scale-Up/Scale-Out) and System Levels
Work closely with the sales and marketing teams to develop product collateral showcasing benchmarks and competitive data
Support ongoing pre-sales evaluations using performance and power models
Experience Qualifications:
Master s/Phd in CS/EE, or a related field, with a solid foundation in Computer Architecture.
At least 7/3 years of experience in performance measurement, benchmarking, modeling, simulation, or system architecture, with a passion for AI technologies.
A strong foundation in AI/Deep-Learning fundamentals to complement your expertise in computer architecture.
Proficient programmer in Python/C/C++.
Experience in building simulators/models in SytemC or similar frameworks at different abstraction levels.
Exceptional problem-solving, analytical, and innovation-driven skills.
Effective communication skills and adeptness at collaborative teamwork.