Sr. Principal Software Scientist

A Moving Experience. Who is Cerence AI?   Cerence AI is the global leader in AI for transportation, specialized in building AI and voice-powered companions for cars, two-wheelers, and more that enable people to focus on what matters most. With over 500 million cars shipped with Cerence AI's technology, we partner with leading automakers (such as Volkswagen, Mercedes, Audi, Toyota and many more), mobility providers, and technology companies to power intuitive, integrated experiences that create safer, more connected, and more enjoyable journeys for drivers and passengers alike.   Our Driving Force    Our team is dedicated to pushing the boundaries of AI innovation, working around the globe with headquarters in Burlington, Massachusetts, USA and 16 other offices across Europe, Asia, and North America. We bring together diverse backgrounds, and varied skill sets with the shared goal of advancing the next generation of transportation user experiences. Our culture is customer-centric, collaborative, fast-paced, and fun, with continuous opportunities for learning and development to support your career growth.   Interested in having a significant impact in a dynamic industry with a high-performing global team? We’re looking for an exceptional Senior   Principal   AI Scientist in Generative AI who is ready to drive the future of mobility with us!   What You Will Work On   Design and train  large ‑ scale transformer and hybrid foundation models   Own model architecture choices across text, multimodal, and emerging paradigms   Diagnose and resolve  training instabilities at scale   Navigate scaling tradeoffs across data, compute, and architecture   Define the technical direction for next ‑ generation models   Core Responsibilities   Deep Learning & Transformer Foundations   Apply strong fundamentals in deep learning and representation learning   Design and modify transformer architectures, including:   Attention variants   RoPE ,  ALiBi   Grouped Query Attention (GQA)   Mixture ‑ of ‑ Experts ( MoE )   Build models  from first principles , not just adapt  pre ‑ existing codebases   Optimi s ation   Dynamics & Training Stability   Own optimi z er and scheduler choices, including:   AdamW   Lion   Adafactor   Learning ‑ rate and warmup schedulers   Understand and debug:   Optimizer instability   Gradient pathologies   Divergence at large scale   Scaling Laws & Compute Tradeoffs   Apply and validate  scaling laws   Navigate  Chinchilla ‑ style compute vs data tradeoffs   Make informed decisions about model size, dataset size, and training duration   Loss Functions & Alignment   Design and experiment with loss functions including:   Next ‑ token prediction   Contrastive objective s   RLHF ,  DPO ,  GRPO   Understand how loss design impacts convergence, generalization, and alignment   Distributed Foundation Model Training   Design and execute large ‑ scale training using:   FSDP   ZeRO ‑ 3   Tensor parallelism   Pipeline parallelism   Apply   Mixed precision ( bf16 ,  fp8 )   Gradient checkpointing   Partner closely with ML systems teams while retaining architectural ownership   Architecture Innovation   Explore and implement novel model designs, including:   MoE  routing strategies   Multimodal fusion architectures   SSM / hybrid architectures   Design architectures with KV cache efficiency and inference implications in mind   What Success Looks Like   Training remains stable as models scale in size and complexity   Architectural decisions are principled and defensible   Models converge faster and generalize better due to architecture and optimi s ation choices   Failure modes are understood, not mysterious   The organization develops true  in ‑ house foundation model expertise   Required Experience & Skills   Strongly Required   Deep theoretical and practical understanding of modern deep learning   Hands ‑ on experience training  large models from scratch   Ability to reason about optimization, not just tune hyperparameters   Comfort operating in ambiguous, research ‑ driven environments   Critical Technical Skills   Transformer internals and attention mechanisms   Optimi s ation   algorithms and training dynamics   Scaling laws and compute/data tradeoffs   Distributed training strategies and mixed precision   Architecture innovation for large, real ‑ world models   Common Problems You’ll Be Solving    Why training diverges at scale   How optimizer dynamics interact with architecture   When scaling laws break down   The real tradeoffs between data, compute, and model design   What we offer   We offer a generous compensation and benefits package (in addition to the base salary), including:   Salary range   $185,000.00 - $280,000.00   It is not typical for offers to be made at or near the top of the range. The actual salary will be determined based on experience and other job-related factors.   Annual bonus opportunity   Insurance coverage (medical, dental, vision, life, and disability)   Paid time off   Paid holidays   Company contribution to the RRSP (Registered Retirement Savings Plan)   Equity awards for certain positions and levels   Remote and/or hybrid work available depending on the position   All compensation and benefits are subject to the terms and conditions of the underlying plans or programs, as applicable, and may be amended, terminated, or replaced from time to time.   Cerence Inc. (Nasdaq: CRNC and  www.cerence.com ) is the global industry leader in creating unique, moving experiences for the automotive world. Spun out from Nuance in October 2019, Cerence is a new, independent company that has quickly gained traction as a leader in the automotive voice assistant space, working with all of the world’s leading automakers – from Ford and Fi