Foundation Models ML Engineer

Amazon; Austin (TX); On-site; USD 143,000 - 193,000; Full time

Benefits offered by this job

Health insurance

401(k) matching

Paid time off

Parental leave

Job summary

Amazon Web Services (AWS) is seeking an AI/ML Engineer to design, train and evaluate foundation models for long-horizon agentic workloads. You will blend model development with systems optimization and hardware-aware techniques.

The role emphasizes efficient, stable model performance, throughput, and memory usage, plus support for multi-step task execution and tool-use in a cloud-native environment. Strong ML/DL background and CUDA/Triton experience preferred.

Qualifications

• Strong background in ML, DL, transformers and foundation models.
• Experience training, fine-tuning or serving large-scale models.
• Familiar with GPU architecture, CUDA, Triton or custom kernels.
• Knowledge of distributed training and model parallelism.
• Experience with AWS accelerators (Trainium/Inferentia) preferred.

Responsibilities

• Design, train, fine-tune, and evaluate custom foundation models.
• Improve efficiency, stability, latency, throughput and memory usage.
• Build models that support planning, tool use, memory, self‑correction and multi-step tasks.
• Optimize training and inference using CUDA, Triton and GPU kernels.
• Profile and resolve compute, memory, and distributed training bottlenecks.
• Work with GPUs and specialized AI accelerators like AWS Trainium.

Skills

Foundation models

Machine learning

Deep learning

Transformers

Distributed training

GPU architecture

CUDA

Triton

Hardware-aware optimization

Model evaluation

Education

PhD in ML/CS/EE

MS in CS/related field

Tools

PyTorch

JAX

CUDA toolkit

Job description

Amazon Web Services (AWS) is seeking an AI/ML Engineer to design, train and evaluate foundation models for long-horizon agentic workloads. You will blend model development with systems optimization and hardware-aware techniques.

The role emphasizes efficient, stable model performance, throughput, and memory usage, plus support for multi-step task execution and tool-use in a cloud-native environment. Strong ML/DL background and CUDA/Triton experience preferred.