Principal Engineer, ML (VLA Automated Driving)

Cariad, Inc.•Mountain View, CA

9h•Hybrid

About The Position

We are CARIAD, an automotive software development team with the Volkswagen Group. Our mission is to make the automotive experience safer, more sustainable, more comfortable, more digital, and more fun. To achieve that we are building the leading tech stack for the automotive industry and creating a unified software platform for over 10 million new vehicles per year. We’re looking for talented, digital minds like you to help us create code that moves the world. Together with you, we’ll build outstanding digital experiences and products for all Volkswagen Group brands that will transform mobility. Join us as we shape the future of the car and everyone around it. Role Summary: The Principal Engineer, ML (VLA Automated Driving) is the technical anchor for Vision-Language-Action (VLA /VLAM) models for our Level 2++ to Level 4 Automated Driving stack. This role defines the technical direction for VLA-based driving across model architecture, training strategy, data flywheel, evaluation, and embedded deployment. The Principal Engineer bridges multimodal foundation-model advances with the realities of real-time, safety-critical automotive systems and helps turn promising research into robust in-vehicle capability. This role also serves as the technical champion for applying GenAI and agentic AI to engineering workflows, identifying and scaling high-value use cases across model development, data flywheel, evaluation, and engineering productivity. This is a highly cross-functional leadership role spanning model, data, evaluation, and deployment.

Requirements

Deep expertise in end-to-end AI and foundation-model approaches for automated driving
Strong software engineering skills and production mindset
Excellent analytical, debugging, and technical decision-making skills
Ability to lead highly complex cross-functional efforts in ambiguous environment
Strong written and verbal communication skills
Ability to collaborate effectively across teams, geographies, and time zones
Deep expertise in foundation models, multimodal learning, and VLA / VLAM approaches for automated driving
Strong background in transformers, vision models, multimodal fusion, and spatio-temporal modeling
Hands-on experience with PyTorch or equivalent ML frameworks
Strong experience developing and adapting large-scale ML models for real-world systems
Experience or strong familiarity with AD/ADAS systems, including end-to-end driving models, world models, or VLA-based architectures
Strong foundation in model evaluation, error analysis, and generalization across diverse driving scenarios

Nice To Haves

Experience with imitation learning, offline RL, reinforcement learning, or world-model-based training
Familiarity with quantization, pruning, distillation, and hardware-aware optimization
Familiarity with TensorRT, ONNX Runtime, or similar inference frameworks
Experience deploying models on embedded or automotive-grade hardware
Experience with simulation and large-scale evaluation pipelines for automated driving
Understanding of automotive system constraints and safety considerations for ML-based ADAS/AD systems

Responsibilities

Define the technical direction for VLA / VLAM-based automated driving
Lead architecture decisions for multimodal driving models from perception and context to trajectories, actions, or driving intent
Drive technical decisions on model adaptation, planning interfaces, fallback/arbitration, latency, and generalization
Design, adapt, and evolve state-of-the-art vision-language-action / multimodal foundation models for automated driving
Define training strategies across supervised learning, imitation learning, offline learning, synthetic data, and simulation-based approaches as appropriate
Drive model quality across robustness, generalization, and complex traffic interactions
Define the data strategy needed to improve model performance quickly and systematically
Partner with data and platform teams to establish a scalable flywheel across data selection, balancing, mining, labeling/annotation inputs, retraining, and evaluation
Align data and training iteration loops to measurable performance outcomes and release readiness
Define evaluation and benchmarking strategies for route-level and scenario-level driving performance
Partner with embedded and systems teams to support deployment on target automotive hardware
Drive model evaluation, error analysis, and generalization assessment across diverse driving scenarios
Serve as a senior technical leader across engineering and partner teams, mentoring others and helping build long-term VLA capability in Mountain View
Act as the technical champion for GenAI and agentic AI workflows within the team, identifying, validating, and helping scale high-value applications across model development, evaluation, data, and engineering productivity
Establish technical standards and best practices for scalable, production-grade ML development in safety-critical systems