Principal AI System Architect and Researcher

Analog Devices•Boston, MA

About The Position

We are building a pioneering organization dedicated to advancing the next generation of Edge AI compute. We are seeking a Principal AI System Architect to take a leading role in shaping our AI workload and SOC Heterogeneous compute architecture. In this position, you’ll dive deep into applications and use cases, helping us identify the most suitable AI workloads and AI model architectures for optimal compute and memory performance. You’ll also drive a proof-of-concept project using digital twin simulation and explore research in SSM and SNN AI workloads for maximum efficiency. Your expertise will help develop a heterogeneous compute solution that integrates various emergent compute modalities. You’ll collaborate closely with the SOC architect, Embedded AI team, and internal and external hardware teams (Analog, RTL Designers) to ensure seamless co-design (hardware and AI), ensuring that compute cores and AI models work together smoothly to deliver state-of-the-art PPAL.

Requirements

PhD degree in electrical engineering, computer engineering, computer science, or a related field.
10+ years of advanced experience in AI model architecture, ANN/SNN/SSM/Sparsity and Codesign AI workload with Compute& Memory
Proficient with multiple frameworks, including Snntorch, PyTorch, TensorFlow, CUDA, C/C++, and Python.
Deep understanding of system-level design, including Analog/RTL, firmware, AI, and tools for embedded and edge deployment.
Skilled at translating complex technical solutions to managers and executives in a clear and intuitive way.

Nice To Haves

3+ years of proven experience in robotics, drones, AI model development, architecture, and simulation (Isaac Sim, AirSim) – Big Advantage

Responsibilities

Translate applications and use cases to AI workloads and model architectures for edge computing, optimizing power efficiency and ensuring high success rates in task completion.
Re-architect AI workloads and co-design solutions using diverse, heterogeneous compute modalities beyond traditional methods.
Drive breakthroughs by converting ANN models (such as CNNs and Transformers) to those that operate efficiently with reduced memory using paradigms like SNN, SSM, sparsity, or RNN.
Create and maintain detailed architecture documentation, concluding all HW requirements for supporting multi-function for end applications.
Drive a proof of concept using a digital twin simulation, and explore ways to leverage open-source AI models and datasets.
Co-design AI models with the most suitable compute paradigms and ensure the coexistence of multiple compute solutions.
Collaborate with startups to explore ways to leverage their IP for ADI needs.
Partner with Embedded AI/ML and System Engineering teams to ensure a successful POC.