Forward Deployed Engineer, Physical AI Infrastructure

Nebius is hiring a Forward Deployed Engineer to work alongside Physical AI customers, shape our core infrastructure products, and test ideas for what we should build next.
You will work with enterprises, ISVs, and digital-native companies running training, simulation, synthetic data, evaluation, and model serving workloads. By understanding their systems and helping them reach production, you will uncover requirements that inform our Cloud, Serverless, and Token Factory teams. The goal is to connect needs across Physical AI industries with product decisions, including improvements that benefit enterprise customers more broadly.
You will also investigate opportunities for new Nebius products. You will develop hypotheses about unmet needs, examine demand and existing alternatives, and build prototypes to test with customers. Your findings will help determine which opportunities deserve investment and what an initial product should deliver.
The work is hands-on. You will write code, deploy and benchmark models, build agents, and troubleshoot infrastructure alongside customer engineers.
The position is part of the Physical AI go-to-market team, with close collaboration across customers, NVIDIA, Product, and Engineering.

• Work with customer engineers to understand their workloads, infrastructure constraints, and production requirements.
• Translate Physical AI use cases into requirements that Cloud, Serverless, and Token Factory teams can use, backed by measurements and customer evidence.
• Identify enterprise adoption barriers and common needs across industries. Work with Product and Engineering to assess gaps and test improvements with customers.
• Build agents that set up, run, and debug Physical AI workloads, connecting them to clusters, schedulers, storage, and model servers.
• Deploy and benchmark open-weight models with serving stacks such as vLLM or SGLang, then tune them for customer workloads.
• Build prototypes and run technical evaluations to answer specific questions about performance, reliability, and product fit.
• Assess demand and existing alternatives for opportunities outside current product teams’ scope, then build where there is a clear market.
• Debug across Python, agents, model servers, Kubernetes, GPUs, networking, and storage. Turn successful setups into reusable scripts, deployments, and short guides.

• Three or more years in software engineering, cloud infrastructure, DevOps, ML infrastructure, or similar, including ownership of a production service, cluster, or platform.
• Experience working directly with customers or external engineers to scope problems, run evaluations, and explain technical tradeoffs.
• Ability to turn an unfamiliar customer problem into concrete infrastructure requirements.
• Practical experience building with LLMs or agents, at work or independently, with an understanding of their internals (e.g., transformers, KV caching, or tool calling) and how to evaluate their outputs.
• Familiarity with enterprise infrastructure requirements, including identity and access management, network isolation, security, monitoring, and reliability.
• Strong Python and Linux skills: you can read unfamiliar code, automate a setup, and debug from logs and metrics.
• Practical experience with containers and Kubernetes, including configuration, networking, storage, and scheduling.
• A record of getting useful software working with incomplete requirements, measuring results, and following through with the people using it.

• Enterprise customer engineering, solutions architecture, consulting, or taking a new product from an initial customer problem to adoption.
• Distributed compute such as Ray, Slurm, Spark, or multi-GPU jobs.
• NVIDIA GPU workloads, including memory management, drivers, CUDA, NCCL, and monitoring.
• Model serving with vLLM, SGLang, TensorRT-LLM, or TGI.
• Robotics, simulation, synthetic data, reinforcement learning, world models, or tools such as Isaac Sim, Isaac Lab, Omniverse, and Cosmos.
• Infrastructure automation and enterprise integrations, including Terraform, CI/CD, private connectivity, and identity federation.