About the job
Exciting opportunity for an Applied Scientist to drive innovation in Generative AI and multimodal LLMs at Adobe. Collaborate with top researchers, prototype and deploy cutting-edge models, and shape the future of AI-powered products. Join a dynamic team and make a real impact on millions of users worldwide.
Responsibilities
Transform innovative research concepts into practical applications within the realm of Generative AI, LLMs, Reinforcement learning, Reasoning, Evaluations.
Prototype and experiment rapidly, demonstrating feasibility and business impact.
Push beyond academic results to develop and deploy practical, differentiated innovations for Adobe’s products.
Collaborate with world-class researchers and ML engineers to bring research ideas to production.
Publish and present your work in world-class scientific venues in AI/ML fields.
Develop and enhance GPU-accelerated pipelines for (customized) model training and inference, focusing on performance, scalability, and reliability.
Foster a culture of innovation, technical excellence, and continuous improvement across the organization.
Qualifications
Minimum
Ph.D. or Masters or equivalent experience in Engineering, Computer Science, AI/ML or related fields and 10+ professional experience.
Research or industry experience in training AI/ML models in at least one of the following modalities: multimodal LLMs, Image, Video.
Proficiency in training and optimizing large-scale models, involving data curation, distributed training, and memory-efficient strategies.
Experience with post-training techniques such as fine-tuning, alignment or distillation.
Proficiency with modern deep learning frameworks (e.g., PyTorch) and experience scaling models on GPU/TPU clusters.
Excellent communication skills and a great teammate.
Preferred
Experience on large-scale generative model training.
Experience on synthetic data generation.
Previous involvement with product teams in technology transfers.
Experience of working with large-scale datasets.
4-7 years of experience in relevant fields.