Staff / Principal Research Engineer - End-to-End Driving (VLA)

Neolix; Sunnyvale (CA); On-site; USD 250,000 - 420,000; Full time

Benefits offered by this job

Founding-team equity

Premium health, dental, vision

401(k)

Job summary

Neolix US is seeking a hands-on technical leader to own the next generation of end-to-end driving using a Vision-Language-Action model family. You will map camera inputs to driving actions, underpin training pipelines and drive validation in real vehicles.

You will lead cross-functional teams, apply state-of-the-art multimodal modeling and ensure scalable, safe integration with vehicle systems across a global program.

Qualifications

• Hands-on experience designing, training and evaluating end-to-end driving, robotics-policy or VLA models mapping raw sensor input to actions.
• Experience at leading autonomous-driving company required.
• Strong knowledge of model training and evaluation with real-vehicle deployments.

Responsibilities

• Lead the design of the end-to-end VLA driving model, including observation and action interfaces, backbone selection, temporal modeling, action representation and training objectives.
• Own the full training chain (pre-training, post-training and closed-loop training) using imitation learning, learning from interventions and reinforcement learning.
• Establish rigorous closed-loop evaluation, with reproducible baselines, ablations, held-out distributions and a behavior failure taxonomy.
• Work with data, training infrastructure, vehicle and safety teams to turn research results into validated vehicle integration.
• Set an ambitious, testable technical roadmap and raise the scientific bar of our global team.

Skills

Autonomous driving

PyTorch/JAX

Multimodal modeling

Temporal modeling

Leadership

Vehicle integration

Education

Technical degree

Tools

PyTorch

JAX

Job description

Neolix US is the Silicon Valley Physical AI research team of Neolix, the global leader in L4 autonomous delivery. Neolix operates 30,000+ autonomous logistics vehicles in 300+ cities across 15+ countries, with ~200 million autonomous kilometers driven on public roads.

About the role

We are looking for a hands-on technical leader to own our next generation of end-to-end driving: a single learned policy, built on a Vision-Language-Action (VLA) model family, that takes camera video and vehicle context as input and outputs driving actions.

Must-have: hands-on experience at a leading autonomous-driving company

(e.g., Waymo, Tesla Autopilot, Wayve, Momenta, XPeng, Cruise or Nuro). General tech experience alone does not qualify.

Key responsibilities

• Lead the design of the end-to-end VLA driving model, including observation and action interfaces, backbone selection, temporal modeling, action representation and training objectives
• Own the full training chain (pre-training, post-training and closed-loop training) using imitation learning, learning from interventions and reinforcement learning
• Establish rigorous closed-loop evaluation, with reproducible baselines, ablations, held-out distributions and a behavior failure taxonomy
• Work with data, training infrastructure, vehicle and safety teams to turn research results into validated vehicle integration
• Set an ambitious, testable technical roadmap and raise the scientific bar of our global team

Qualifications

• Required: experience at a leading autonomous-driving company; general tech experience alone does not qualify
• Hands-on experience designing, training and evaluating end-to-end driving, robotics-policy or VLA models that map raw sensor input to actions
• Deep expertise in PyTorch or JAX, multimodal and temporal modeling, and adapting large pre-trained models to action prediction
• A track record of connecting model research to closed-loop evaluation on a real vehicle, robot or deployed system
• Working knowledge of vehicle integration constraints, including action interfaces, physical feasibility, latency and memory budgets
• Proven technical leadership across research, systems, data and integration teams
• A relevant technical degree or equivalent practical experience; a PhD is not required

Preferred

• Training VLA or vision-action policies at scale; action tokenization, diffusion or flow-based action heads; distillation to onboard compute
• Offline or online reinforcement learning, with a focus on generalization and recovery
• Sensor-level closed-loop simulation, neural reconstruction or action-conditioned world models for evaluation
• Relevant publications, open-source contributions or proven production impact
• Equity: generous founding-team equity package
• Benefits: premium health, dental and vision insurance, 401(k), and comprehensive Silicon Valley tech perks

This is a founding technical-ownership role. You will define a core learned-driving capability end to end, combining frontier modeling with the real-world demands of autonomous delivery.