About this role
Sr. Principal Software Scientist at Cerence. Location: United States. Role: designing models, training models, diagnosing instabilities Requirements: Deep expertise in deep learning and foundation models; hands-on experience training large models from scratch; strong optimization, distributed training, and architecture innovation skills. Category: Research and Development (R&D) Seniority: Senior Level Tools: FSDP, ZeRO-3, AdamW, Lion, Adafactor, bf16, fp8, RoPE, ALiBi, Grouped Query Attention (GQA), MoE, RLHF, DPO, GRPO, Tensor parallelism, Pipeline parallelism, Gradient checkpointing, Chinchilla Commitment: Full Time Workplace: Remote Languages: English