Capacity Path

Job Summary:
Netflix is a company on a mission to entertain the world by merging creativity and cutting-edge technology. The Inference Specialist, Creative Technology will support the Production, Research, and Engineering teams by executing model inference workflows and translating creative needs into reliable production practices.
Responsibilities:
• Operate and support custom generative AI inference workflows across a wide variety of film and series projects
• Run, monitor, and troubleshoot GPU-based inference jobs across local workstations, cloud infrastructure, and/or cluster environments, including distributed multi-GPU runs
• Prepare and validate inputs for model inference, including video, image, audio, masks, conditioning assets, prompts, metadata, and configuration files
• Tune inference parameters in collaboration with Creative Technology leadership, artists, researchers, and engineers to achieve production-quality results
• Debug failed or degraded runs by inspecting logs, outputs, configs, model checkpoints, data shapes, masks, frame ranges, codecs, GPU utilization, and environment issues
• Maintain clean, repeatable inference launch workflows, including scripts, config templates, run manifests, output naming conventions, and result tracking
• Partner with researchers and engineers to test new models, checkpoints, samplers, conditioning methods, and pipeline changes in real production scenarios
• Translate experimental model capabilities into usable production practices
• Identify friction in inference workflows and drive improvements through tooling, automation, documentation, and better defaults
• Support rapid iteration with artists and creative stakeholders by preparing outputs for review, comparing variations, tracking parameters, and surfacing clear recommendations
• Own quality control for generated outputs
• Help bridge communication between creative, production, research, and engineering teams by explaining technical constraints and creative tradeoffs clearly
• Maintain awareness of GPU capacity, queue status, runtime expectations
• Contribute to a culture of practical experimentation: move quickly, test carefully, document learnings, and turn one-off fixes into repeatable workflows
Qualifications:
Required:
• 4+ years of relevant experience in machine learning production, VFX technology, post-production engineering, creative technology, technical direction, or a closely related technical production role
• Hands-on experience running GPU-based model inference for image, video, audio, or multimodal generative AI systems
• Experience working with Python-based ML codebases and command-line workflows in Linux environments
• Experience debugging production runs using logs, stack traces, configuration files, model inputs, and generated outputs
• Working knowledge of deep learning inference concepts, including checkpoints, schedulers or samplers, seeds, precision, batching, conditioning, and GPU memory constraints
• Experience with video and image production formats, including frame sequences, ProRes, H.264/H.265, EXR, PNG, MP4/MOV containers, resolution handling, frame rates, and colorspace considerations
• Experience coordinating technical work across creative, production, research, and engineering stakeholders
• Demonstrated ability to operate effectively in a fast-moving R&D environment where tools, models, and workflows change frequently
• Strong practical understanding of generative AI inference workflows, especially for video, image, audio, or multimodal models
• Comfort working in Linux shells, Python environments, Git repos, config files, logs, and GPU infrastructure
• Strong debugging instincts: able to isolate whether a problem is data, model, environment, code, infrastructure, or user configuration
• Ability to reason about video and tensor fundamentals, including frame counts, aspect ratios, spatial resolution, temporal alignment, masks, channels, and batch dimensions
• Experience with tools and libraries commonly used in production ML workflows, such as PyTorch, CUDA, ffmpeg, OpenCV, NumPy, safetensors, and distributed launch tools
• Comfort with job schedulers, cloud GPU environments, or cluster workflows; Slurm experience is a strong plus
• Careful eye for generated output quality, including temporal artifacts, mask errors, motion issues, color shifts, compression problems, and sync problems
• Able to balance creative iteration speed with technical rigor, reproducibility, and clear communication
• Self-directed and ownership-minded; comfortable seeing a messy problem, creating a path through it, and pulling in help when needed
• Collaborative and calm under pressure, especially when supporting time-sensitive creative reviews or production deadlines
• Strong written communication, including the ability to document workflows, summarize test results, and explain technical findings to non-technical partners
• Comfort with ambiguity, rapidly changing tools, and incomplete information
• Genuine interest in tooling for filmmakers, with the curiosity to engage deeply with both the creative possibilities and the engineering realities of the work
Company:
Netflix is an online streaming platform that enables users to watch TV shows and movies. Founded in 1997, the company is headquartered in Los Gatos, USA, with a team of 10001+ employees. The company is currently Late Stage.