Embedded AI Engineer: Generative Models for Clients

Amazon Web Services (AWS); East Palo Alto (CA); On-site; USD 144,000 - 214,000; Full time

Benefits offered by this job

Health insurance

401(k) matching

Paid time off

Parental leave

Job summary

Amazon Web Services, Inc. seeks a Forward Deployed Engineer II to embed with strategic customers, building and optimizing custom generative AI models in production.

You drive end-to-end model workstreams from evaluation to deployment, collaborating with customer engineering teams. You will customize, post-train models, build data pipelines, and develop large-scale training/inference systems on AWS Trainium, aiming for optimal latency, throughput and cost.

Qualifications

• 3+ years of non-internship software development experience.
• 2+ years designing or architecting scalable systems.
• Experience programming in at least one language.
• Hands-on experience with generative AI technologies (LLMs, vision-language, or speech models).

Responsibilities

• Embed with customer ML and engineering teams to translate goals into a plan and measurable evaluation bar.
• Customize and post-train models, build data pipelines, and evaluate model improvements.
• Design distributed training and inference pipelines and optimize latency and cost.
• Migrate and tune workloads for AWS Trainium; benchmark accuracy vs GPU baselines.
• Codify reusable fine-tuning recipes and contribute to AWS service teams with learnings.

Skills

Professional software development

System design

Programming in at least one language

Generative AI experience

Education

Bachelor's degree in computer science or equivalent

Tools

PyTorch FSDP2

Megatron-LM

Miles

OpenEnv

SGLang

vLLM

NeMo

Job description

Amazon Web Services, Inc. seeks a Forward Deployed Engineer II to embed with strategic customers, building and optimizing custom generative AI models in production.

You drive end-to-end model workstreams from evaluation to deployment, collaborating with customer engineering teams. You will customize, post-train models, build data pipelines, and develop large-scale training/inference systems on AWS Trainium, aiming for optimal latency, throughput and cost.