Applied Scientist
Core
Develop AI innovations from research concepts through product validation and technology transfer, focusing on image and video editing.
Role type
Applied Scientist (Multimodal AI)
Builds
Multimodal architectures (vision encoders, modality projectors, cross-modal attention) for Adobe products like Firefly and Creative Cloud.
Domain
Creative software, AI, Multimodal Machine Learning
Deliverable
production ML models
Required skills
Transformer architecture development, Diffusion model development, Multimodal model training, Python, PyTorch, End-to-end model development, Post-training techniques (SFT, RL, alignment), Model evaluation and failure analysis
Preferred skills
Large-scale distributed training, Inference optimization (latency, throughput, cost), Synthetic data generation and quality control
Technologies
PyTorch, Python, Transformers, Diffusion models
Responsibilities
Identify research advances for product differentiation, Design and optimize multimodal architectures, Build rigorous evaluation pipelines for quality and safety, Diagnose model failures through ablation studies and error analysis
Seniority
Early-career to Senior, hands-on IC
