דלג לתוכן הראשי

Research Engineer, Multimodal Systems

Technology
חברה בתחום הייטק / חומרה / תוכנה / סייבר
תל אביב יפו, ישראללפני 2 שבועות
משרה מלאה

תיאור המשרה

Required Research Engineer, Multimodal SystemsBuild the training and inference backbone behind our image-to-design foundation model. 5+ years of engineering experience required.About the role:Great research dies without great engineering. You will be the person who makes big training runs fast and reproducible, turns a promising checkpoint into a sub-60-second product experience, and builds the data and eval infrastructure that lets our researchers move quickly. You'll work across the whole stack, from distributed training jobs to the low-latency inference path our Figma plugin calls in real time.In this role you will:Own the training infrastructure: multi-node GPU orchestration, data loading at scale, checkpointing, experiment tracking, and profiling.Build the inference pipeline that serves our models to designers with the latency and reliability the product demands.Engineer the data flywheel: ingestion, labeling, dedup, and eval sets that turn real usage into training signal.Partner with research to prototype quickly and harden what works into production.Requirements: You may be a good fit if you haveAt least 5 years of professional software engineering experience, including production ML or large-scale distributed systems.Strong Python and PyTorch, and fluency with GPU training and serving.Hands-on experience with distributed training and cloud GPU infrastructure (AWS, GCP, Kubernetes, Ray, or similar).A bias for shipping and an allergy to flaky pipelines.

You make things fast, observable, and reproducible.Nice to have:CUDA or Triton kernels inference optimization (batching, KV-cache, quantization) vision or multimodal model experience building internal eval and data tooling from scratch.This position is open to all candidates.

Keywords
Plug-inOrchestrationCUDAPyTorchFigmaPython

מתעניינים במשרה הזו?

משרות דומות