AI Engineer, Recursive Self-Improvement for Compute

WHAT YOU DO AT AMD CHANGES EVERYTHING

At AMD, our mission is to build great products that accelerate next‑generation computing experiences—from AI and data centers, to PCs, gaming and embedded systems. Grounded in a culture of innovation and collaboration, we believe real progress comes from bold ideas, human ingenuity and a shared passion to create something extraordinary. When you join AMD, you’ll discover the real differentiator is our culture.

We push the limits of innovation to solve the world’s most important challenges—striving for execution excellence, while being direct, humble, collaborative, and inclusive of diverse perspectives. Join us as we shape the future of AI and beyond.

Together, we advance your career.

We are hiring AI Engineers to build recursive self‑improvement systems for compute. This role sits at the intersection of AI systems, performance engineering, hardware‑aware optimization, and agentic software development. You will help build systems where AI proposes improvements, verifies correctness, measures impact, learns from failures, and improves the next generation of compute workloads and platforms.

The work requires turning complex engineering tasks into repeatable optimization loops with clear inputs, candidate generation, automated validation, measurable scoring, and reliable iteration. You will work on problems where feedback may be expensive, correctness is non‑negotiable, and small improvements can have large impact at scale.

You are a hands‑on engineer who can move between modern AI methods, systems performance, and hardware‑aware workflows. You are comfortable with fast‑moving AI research, but you care most about measurable outcomes: correctness, speed, efficiency, quality, reproducibility, and engineering leverage. You can collaborate with research scientists, hardware engineers, software engineers, and external technical partners while keeping an end‑to‑end loop working.

• Build agentic and learning‑driven optimization loops for compute workloads and hardware engineering workflows.
• Develop systems that generate, compile, test, benchmark, profile, and iterate on candidate improvements with minimal human intervention.
• Convert high‑value engineering workflows into verifiable tasks with clear graders, harnesses, metrics, and failure feedback.
• Collaborate with AI researchers on reward design, reward shaping, reward hacking analysis, long‑horizon optimization, and model improvement loops.
• Design feedback systems that accumulate useful data from successful attempts, failed attempts, profiler traces, benchmark results, and validation logs.
• Improve iteration speed through staged validation, caching, parallel execution, proxy metrics, and faster feedback paths.
• Build reusable tools and patterns that can generalize across multiple compute and hardware optimization domains.
• Mentor other engineers and help set technical direction for self‑improving AI systems for compute.

• AI‑assisted program optimization, code generation, debugging, and automated repair.
• GPU and CPU performance engineering, including kernels, libraries, compilers, profilers, and benchmark‑driven optimization.
• Agentic workflows that use tools, tests, simulators, profilers, and structured feedback to improve over time.
• Reinforcement learning, post‑training, reward modeling, and evaluation methods for engineering tasks.
• Hardware‑aware optimization where correctness, latency, throughput, area, power, timing, or resource usage may all matter.
• Experiment platforms, leaderboards, dashboards, and data pipelines for reproducible comparison and continuous improvement.

• Strong software engineering experience in Python and at least one systems language such as C++, C, HIP, CUDA.
• Experience building AI, ML, agentic, optimization, or automation systems that are evaluated with objective metrics.
• Ability to design reliable experiment loops, benchmark harnesses, validation workflows, and correctness/performance evaluation pipelines.
• Strong technical judgment in debugging, profiling, root‑cause analysis, and performance‑oriented iteration.
• Clear communication…