Senior AI scientist

Overview

Senior AI Scientist to work on Cerence’s conversational AI platform using large language models, including multimodal models for edge and cloud.

What you'll do

• Design and train large-scale transformer and hybrid foundation models.
• Diagnose and resolve training instabilities at scale.
• Navigate scaling tradeoffs across data, compute, and architecture.
• Define the technical direction for next-generation models.
• Design and modify transformer architectures (e.g., attention variants and MoE).
• Own optimizer and scheduler choices and debug training dynamics.
• Design and execute large-scale distributed foundation model training.

What you'll need

• Strong fundamentals in deep learning and representation learning.
• Experience designing/ modifying transformer architectures including attention variants (RoPE, ALiBi, GQA) and Mixture-of-Experts (MoE).
• Knowledge of optimizers and training components including AdamW, Lion, and Adafactor.
• Understanding of learning-rate and warmup schedulers.
• Ability to apply and experiment with loss functions such as next-token prediction and contrastive objectives.
• Familiarity with RLHF and objectives including DPO and GRPO.
• Experience with distributed training using FSDP, ZeRO-3, tensor parallelism, and pipeline parallelism.