VLA Engineer

Department: ENGINEERING

Location: Pangyo (Software Dream Center), South Korea

Employment Type: FullTime

We are looking for the best

About Us

At 42dot, we build multimodal driving models that connect visual perception, scene context, and driving actions to enable robust autonomous driving behavior.
Our goal is to advance planning, decision-making, and action generation in real-world driving environments by leveraging VLA-style modeling approaches that incorporate prior knowledge for improved generalization, with a strong focus on production deployment and on-vehicle execution.

Responsibilities

• Design and develop VLA-based models for autonomous driving, focusing on planning, decision-making, and action generation

• Apply imitation learning, reinforcement learning, and generative modeling to improve driving behavior and long-horizon decision-making

• Fine-tune and adapt multimodal models (VLM / VLA-style architectures) for autonomous driving tasks using large-scale driving datasets

• Define and evaluate multimodal representations for driving data (video, BEV, map, vehicle state, actions, optional language annotations)

• Build and maintain end-to-end machine learning pipelines from data curation and training to evaluation and deployment

• Evaluate models in open-loop, closed-loop simulation, and real-vehicle environments, with a focus on safety and robustness

• Collaborate with perception, prediction, planning, control, and platform teams to integrate ML models into production vehicle software stacks

Qualifications

• Strong hands-on experience with deep learning models for autonomous driving, robotics, or sequential decision-making systems

• Practical experience applying imitation learning and/or reinforcement learning to real-world problems

• Solid understanding of Transformer-based architectures and multimodal learning

• Proven experience deploying machine learning models in production or safety-critical systems

• Strong programming skills in Python and experience with PyTorch

• Experience working with large-scale datasets and distributed training environments

• Ability to collaborate effectively across software, vehicle, and hardware teams

Preferred Qualifications

• Experience with VLM / VLA-style models applied to autonomous driving or robotics

• Experience with closed-loop simulation, SIL/HIL, or real-vehicle testing

• Experience optimizing inference using TensorRT, CUDA, quantization, or pruning

• Experience deploying models on embedded or vehicle-grade hardware

• Research or engineering contributions in autonomous driving, robotics, or machine learning

Interview Process

• Resume Screening - Coding Test - Virtual Interview (approximately 1 hour) - Onsite or Virtual Interview (approximately 3 hours) - Final Offer

• Please note that the interview process may vary depending on the position and is subject to change based on scheduling and other circumstances.

• Interview schedules and results will be communicated individually via the email address provided in your application.

Additional Information

• Please upload all required documents in PDF format.

• Veterans and applicants eligible for employment protection will receive preferential consideration in accordance with applicable laws and regulations.

• In compliance with the Act on Employment Promotion and Vocational Rehabilitation for Persons with Disabilities, registered individuals with disabilities will receive preferential consideration.

• 42dot does not accept unsolicited resumes from search firms. We will not pay any fees for resumes submitted without prior agreement.

• A 3-month probationary period may apply.

※ Please make sure to review the information below before applying.

• Learn more about how we work at 42dot, 42dot Way →

• Explore 42dot’s unique Employee Engagement Program, Employee Engagement Program →