Engineer I, VLA Model Development (Intern)

Cariad•Mountain View, CA

20h•Onsite

About The Position

The Engineer I, VLA Model Development (Intern) supports research and development of models for autonomous driving at CARIAD Inc. and the Volkswagen Group. Working with researchers and engineers, the intern develops and evaluates vision-language-action (VLA) models that connect visual observations and language with driving actions. Depending on project needs and research strengths, the work may focus on perception, scene understanding, or world models that predict how driving environments evolve. The position involves prototyping, training, and fine-tuning models or supporting components; preparing driving datasets; running reproducible experiments; evaluating model performance, robustness, driving behavior, and inference cost; analyzing failures in offline and simulation evaluation; and collaborating with planning and deployment engineers on model interfaces. The intern documents methods, results, and limitations, reviews code, presents findings, and delivers a reproducible prototype and final technical report. This is an on-site internship five days per week in Mountain View, California. The candidate must be local; relocation is not available. Immediate availability is required, with a prompt start following offer acceptance and successful completion of a standard background check and drug screening. Work arrangement and internship schedule are TBD. It is a W-2 hourly position. The posting gives an hourly range of $44.88–$96.00 in one section and $44.88–$69.00 for Mountain View in the compensation section; compensation depends on factors including geography, credentials, experience, qualifications, and training. CARIAD is an automotive software development team within the Volkswagen Group, building a unified software platform and digital products for Volkswagen Group brands. The company describes its mission as making the automotive experience safer, more sustainable, more comfortable, more digital, and more fun. Employment is subject to screening, including export-control and sanctions compliance; it may depend on verification of U.S. citizenship or lawful permanent resident status, or confirmation that a required license, exemption, or exception applies.

Requirements

Strong Python programming skills and practical experience with PyTorch or an equivalent deep learning framework.
Foundations in deep learning, linear algebra, probability, and experimental design.
Hands-on experience in at least one area: vision-language-action models, computer vision/perception, or world models.
Ability to prepare datasets, train or fine-tune models, debug failures, and evaluate results reproducibly.
Research, coursework, or project experience developing and evaluating deep learning models; prior industry experience is not required.
Currently pursuing or holding a master’s or PhD degree in Computer Science.
Must be local to Mountain View, California; relocation is not available.
Immediate availability is required.
Successful completion of a standard background check and drug screening is required.
Must participate in export-control and sanctions compliance screening; employment may be contingent on citizenship or lawful permanent resident verification or applicable authorization.

Nice To Haves

Experience with transformers, multimodal learning, imitation learning, diffusion models, or action-conditioned prediction.
Experience with driving video, 3D, scene understanding, multi-sensor data, or simulation.
Familiarity with GPU training, distributed experiments, or model inference optimization.
Research publications, open-source contributions, or a substantial research project.
Currently pursuing a PhD in Computer Science, particularly with research in machine learning, computer vision, or embodied AI.

Responsibilities

Prototype, train, and fine-tune VLA models or supporting perception and world-model components.
Investigate recent methods and implement a focused research project with mentor guidance.
Prepare driving datasets and evaluation splits and improve data quality and coverage.
Run controlled experiments and ablations; maintain reproducible configurations and experiment records.
Measure model accuracy, driving behavior, robustness, and inference cost.
Analyze failures in offline and simulation evaluation.
Collaborate with planning and deployment engineers on model interfaces.
Document methods, results, and limitations; review code and present findings.
Deliver a reproducible prototype and final technical report.