Forward Deployed Engineer (Model Customization and Optimization), AWS Forward Deployed Engineering
Core
Embed with strategic customers to build, optimize, and run custom generative AI models in production, owning workstreams from data preparation to inference optimization.
Role type
Senior IC machine-learning engineer (model customization and optimization)
Builds
Customized open-weight and foundation models, distributed training/inference pipelines, and production inference stacks for AWS Trainium.
Domain
Generative AI, Large Language Models (LLMs), Cloud Infrastructure (AWS)
Deliverable
production ML models
Required skills
Generative AI (LLMs, vision-language, speech), Model customization (pre-training, SFT, DPO, RLHF, agentic RL), Distributed training frameworks (PyTorch FSDP2, Megatron-LM, NeMo), Inference optimization (SGLang, vLLM, quantization), AWS Trainium/Neuron SDK, Data pipeline construction
Preferred skills
Large-scale distributed training on multi-node clusters, LLM post-training frameworks (Miles, verl, OpenRLHF), Custom kernel development (CUDA, Triton), Production ML operations in customer environments
Technologies
PyTorch, Megatron-LM, NeMo, SGLang, vLLM, AWS Trainium, AWS Neuron SDK, AWS SageMaker AI, CUDA, Triton
Responsibilities
Embed within customer ML teams to understand business goals and operational constraints; Customize and post-train models for new domains and modalities; Design and implement distributed training and RL pipelines; Optimize training and inference workloads for latency, throughput, and cost on AWS chips; Codify reusable recipes and tools to accelerate future deployments.
Seniority
Senior, hands-on IC