Principal Machine Learning Engineer, Foundation Models

Cambridge Mobile Telematics (CMT) is the world’s largest telematics service provider with a mission to make the world’s roads and drivers safer. The company’s AI-driven platform, Drive Well Fusion, gathers sensor data from millions of IoT devices — including smartphones, proprietary Tags, connected vehicles, dashcams, and third-party devices — and fuses them with contextual data to create a unified view of vehicle and driver behavior.

Auto insurers, automakers, commercial mobility companies, and the public sector use insights from CMT’s platform to power risk assessment, safety, claims, and driver improvement programs. Headquartered in Cambridge, MA, with offices in Budapest, Chennai, Seattle, Tokyo, and Zagreb, CMT measures and protects tens of millions of drivers across the world every day.

We are embarking on a transformative journey with Drive Well Atlas, a groundbreaking initiative to build a family of novel AIs on telematics data. As a Principal Data Scientist on the Drive Well Atlas team, you will be at the forefront of developing these next-generation AIs. You will lead innovative projects focused on designing, pre-training, fine-tuning, and deploying these AIs. Your work will directly contribute to enhancing our capabilities in risk assessment, driver engagement, and crash and claims processing.

This role requires a deep understanding of modern AI methods, machine learning, physics-informed modeling, and experience with sensor data.

Responsibilities

• Lead the design, pre-training, fine-tuning, and deployment of novel AIs for telematics applications.
• Develop and implement novel algorithms that represent physical phenomena associated with vehicle and human movements.
• Pioneer advanced self-supervised learning techniques, including the design and implementation of innovative tasks tailored to multi-modal telematics sensor data to learn rich representations of movement and driver behavior.
• Drive the development of models robust to noise, missing data, and diverse operating conditions typical of real-world mobile sensor and IoT datasets.
• Build and manage scalable training and inference pipelines using tools like Ray, PyTorch DDP, Horovod, or similar frameworks.
• Integrate these AIs into production systems while ensuring high performance and reliability.
• Optimize these AIs for efficient deployment on various platforms, including cloud and edge/mobile devices.
• Collaborate closely with engineering, product, and research teams to translate cutting-edge research into impactful products and features for the Drive Well Atlas platform.
• Mentor junior scientists and contribute to the broader AI/ML strategy at CMT.
• Stay abreast of the latest AI advancements, evaluating and adopting emerging technologies and methodologies relevant to telematics.
• Contribute to efforts in AI explainability and interpretability.
• Complete any tasks as they arise

Qualifications:

• PhD or MS in Artificial Intelligence, Computer Science, Electrical Engineering, Physics, Mathematics, Statistics, or a related field.
• 7+ years of experience in AI/ML.
• 3+ years of hands-on experience developing and deploying foundation models (e.g., BERT, GPT, and/or custom domain-specific transformers), with a strong portfolio in generative AI for sequential or spatio-temporal data.
• Strong, hands-on experience in building and training multi-modal and/or time-series transformer architectures for complex sensor fusion and behavioral modeling tasks is required.
• Deep expertise in designing pretext tasks for self-supervised learning on noisy, real-world sensor data.
• Proficiency in Python and common data science libraries (e.g., Pandas, Num Py, scikit-learn).
• Extensive experience with deep learning frameworks such as PyTorch (preferred) or Tensor Flow for large-scale model training and deployment.
• Solid understanding and practical experience with distributed training techniques and efficient training methodologies for large models.
• Experience building and maintaining large-scale data processing pipelines and machine learning infrastructure using tools like Spark, Airflow, Docker, and cloud platforms (e.g., AWS, GCP, Azure).
• Excellent problem-solving skills and the ability to…