This role pays more than
86%
of similar roles. Most pay
$188,200–$254,750
— the shaded band above.
At the midpoint, this role pays about
$292k
versus about
$221k
for comparable roles.
Based on 239 similar postings.
Employer
About Amd
AMD (Advanced Micro Devices) is a semiconductor company that develops high-performance processors, graphics cards, and adaptive computing solutions for gaming, data centers, and embedded markets. Industry: Semiconductors
Amd currently has
367 open roles
on FindRole.
Listed pay typically runs
$166,400–$249,600
across 367 roles with salary data.
The Agentic AI Systems Architecture Fellow joins the Software and Solutions Team to define the architectural vision and technical strategy for next-generation Agentic AI systems. This individual contributor role focuses on building end-to-end reference architectures that leverage AMD’s heterogeneous computing portfolio, including CPUs, GPUs, DPUs, networking, and software platforms. The Fellow will develop real-world use cases, proof-of-concepts, and orchestration frameworks to optimize task decomposition, reasoning pipelines, and data-centric performance across both AMD+AMD and AMD+NVIDIA environments. Key technical requirements include expertise in Agentic AI systems architecture, heterogeneous AI infrastructure, and AI performance engineering. The role utilizes technologies such as ROCm™, AMD Zen Software Studio, and AMD Enterprise AI while integrating frontier foundation models with RAG and sglang. The work addresses the challenge of creating secure, scalable, and deterministic AI solutions for enterprise, cloud, and edge deployments.
Define the architectural vision and long-term technical strategy for Agentic AI systems on AMD's heterogeneous computing portfolio.
Develop reference architectures that optimize execution across CPUs, GPUs, DPUs, networking, memory, and storage.
Create real-world use cases and proof-of-concepts to demonstrate AMD platform capabilities to customers.
Architect multi-agent systems and orchestration frameworks for task decomposition, workflow scheduling, and reasoning pipelines.
Design data optimization strategies and integrate frontier foundation models with retrieval systems and enterprise applications.
Improve deterministic AI behavior by incorporating security practices like data provenance and trusted execution.
Influence AMD’s silicon and software roadmaps through collaboration with engineering teams and industry partners.
Serve as a technical authority by representing the company at conferences, in publications, and during customer engagements.
What we're looking for
Must be a recognized industry expert in Agentic AI systems architecture, heterogeneous AI infrastructure, and AI performance engineering.
Experience with agentic AI benchmarks including agent orchestration, planning, routing, and verification is required.
Ability to demonstrate the advantages of EPYC-based platforms for agentic workloads compared to other x86 and ARM alternatives.
Experience designing large-scale AI infrastructure or distributed systems across CPUs, GPUs, DPUs, networking, and storage (preferred).
Understanding of modern AI software ecosystems including ROCm, frontier foundation models, LLM inference, RAG, and orchestration frameworks (preferred).
Expertise in data optimization, model deployment, and supporting enterprise-scale AI systems (preferred).
Strong knowledge in AI performance engineering regarding latency, throughput, token efficiency, and resource utilization (preferred).
Master’s degree in Computer Science, Computer Engineering, Electrical Engineering, AI, Data Science, or a related technical discipline (preferred).
Agentic AI
ROCm
Python
C++
C
LLM Inference
RAG
Vector Databases
Distributed Systems
Heterogeneous Computing
Confidential Computing
Cloud Infrastructure
Performance Engineering
Data Provenance
Networking
Virtualization
Agentic AI
LLMs
Generative AI
RAG
Python
Kubernetes
Docker
CI/CD
LangChain
LlamaIndex
NVIDIA NeMo
Triton
Multi-agent Orchestration
API Gateways
Distributed Systems
Linux
Text-to-SQL
Model Serving
Observability