A

Amazon

Sells products online and rents cloud computing

Senior Forward Deployed Engineer - Model Customization and Optimization

Natural Language Processing & LLMsConfirmed live in the last 24 hours
Full-time

About the Job

Requirements

  • Five or more years of non-internship professional software development experience.
  • Five or more years of programming experience with at least one software programming language.
  • Five or more years of experience leading the design or architecture of new and existing systems, including design patterns, reliability, and scaling.
  • Experience as a mentor, technical lead, or engineering team lead.
  • Hands-on experience training, fine-tuning, or optimizing deep learning models.
  • Experience communicating technical concepts to technical and non-technical audiences, including senior leadership.

Responsibilities

  • Lead customer engagements as the technical lead for model customization and optimization programs spanning multiple workstreams; partner with customer engineering and product leadership to frame business problems, define success metrics and evaluation criteria, and set the technical roadmap from proof of value through production; communicate trade-offs, risks, and results to executives and engineers.
  • Determine the customization strategy to meet quality, latency, and cost targets, including prompting, distillation, supervised fine-tuning, preference optimization, reinforcement learning, continued pre-training, task-specific adapters, or building domain foundation models; own data strategy, synthetic data generation, and evaluation methodology, and drive solutions into production.
  • Design distributed training and reinforcement-learning post-training systems for dense and mixture-of-experts models using frameworks such as PyTorch FSDP2, Megatron-LM, Miles, and verl, and production inference architectures using SGLang and vLLM.
  • Architect reinforcement-learning interaction harnesses using NeMo Gym and OpenEnv, defining connections between agent workflows, tools, environments, reward functions, training, and evaluation.
  • Set parallelism, precision, and serving strategies to balance throughput, latency, memory, reliability, and cost per token, ensuring designs withstand production load.
  • Lead migration of training and inference workloads to AWS Trainium using the AWS Neuron SDK; diagnose performance and accuracy-parity gaps against GPU baselines, direct kernel-level optimization, and work with AWS Neuron and service teams to close capability gaps.
  • Translate benchmark evidence into recommendations that inform customer infrastructure and capacity decisions.
  • Provide technical leadership to FDEs and scientists on engagements, break ambiguous problems into workstreams, and unblock others; mentor FDE IIs on model systems design and production ownership; lead design reviews, provide code reviews, and participate in hiring and interviewing.
  • Develop reusable fine-tuning recipes, evaluation frameworks, benchmarking tools, and reference architectures from patterns across engagements.
  • Influence AWS service roadmaps, including Amazon Bedrock, Amazon SageMaker AI, and AWS Neuron, using field evidence; identify and shape opportunities to expand model customization and optimization within customers and across similar customers.
  • Own engagement delivery plans, set milestones, track progress against deadlines, ensure deliverable quality and completeness, hold the team accountable for customer commitments, escalate risks, and adjust scope or resources when timelines are at risk.

Desired Qualifications

  • A bachelor's degree in computer science or equivalent.
  • Two or more years of experience deploying and operating machine-learning models in customer production environments, working with teams that build and scale end-to-end production systems.
  • Experience defining evaluation strategies and building evaluation frameworks or benchmarks for generative models.
  • Experience leading model training or inference optimization for latency, throughput, or cost using serving engines such as SGLang, vLLM, or TensorRT-LLM; AWS Trainium with Neuron SDK; AWS services such as SageMaker AI, SageMaker HyperPod, or EKS; or custom kernels using CUDA, Triton, or NKI.
  • Experience leading large language model post-training, including supervised fine-tuning, direct preference optimization, reinforcement learning from human feedback, reinforcement learning with verifiable rewards, or multi-turn agentic reinforcement learning, using frameworks such as Miles, slime, verl, NeMo-RL, OpenRLHF, SkyRL, or TRL.
  • Experience leading large-scale distributed training of large language models, including mixture-of-experts models, on multi-node accelerator clusters using PyTorch FSDP2, TorchTitan, Megatron-LM or Megatron Core, NeMo, DeepSpeed, or NeuronX Distributed.

About the Company

Amazon

Sells products online and rents cloud computing

Amazon runs the largest online store in the United States and rents cloud computing to businesses and governments through Amazon Web Services. AWS pioneered renting computing power over the internet, and Amazon built its own delivery network to get orders to customers quickly. Shoppers buy from Amazon and from independent sellers on its store, and Prime members pay a subscription for fast free shipping, streaming video and music, and other perks. AWS rents out computing power, storage, databases, and AI tools so companies can run software without owning servers, and Amazon also makes Kindle e-readers and Echo speakers with the Alexa assistant. Jeff Bezos founded Amazon in 1994, and it is based in Seattle, Washington.

Company size
10,001+
Company stage
IPO
Headquarters
Seattle, WA
Founded
1994

Recently Posted Jobs

Explore more opportunities at Amazon

See more jobs