Deep Learning Researcher
Posted Updated
Description
Tensor Yapay Zeka in Istanbul seeks a researcher to design and train a self-supervised foundation model for multi-channel imagery and time series, focusing on masked modeling and JEPA. You will lead large-model training with PyTorch, DDP/FSDP, mixed precision, and experiment tracking; develop data pipelines, evaluate with probes, and adapt representations to low-label tasks for product integration.
Ideal candidates bring PhD (or equivalent) in CE/CS, hands-on self-supervised learning, and
Design and train a self-supervised foundation model for multi-channel imagery and time series Make architecture, objectives and scaling decisions through experiments Build the multi-GPU training setup and data pipeline for continuously arriving data Evaluate with linear probes, ablations and leakage-free validation Adapt learned representations to low-label tasks for product integration BS/MSc/PhD in CE&CS or equivalent experience Large-model training experience in PyTorch with DDP or FSDP Hands-on self-supervised learning and vision transformers Good use of LLM agents for coding, with awareness of risks and checks Small, fast teamSkills
Sign in to read and join the discussion.