Sr. Simulation/Evaluation Learning Algorithm Engineer (Multi-Camera Vision Simulation/Evaluation)

• Large-Scale Learning and Evaluation Pipeline Construction: Build a large-scale simulation-based learning and optimization environment to enhance algorithm performance. Utilize vast amounts of data generated from simulations to train deep learning components within modules, such as the cost function of a path planner or the adaptive parameters of a controller, and optimize classical algorithm parameters through iterative, automated testing.
• High-Fidelity Simulation Platform Operation: Design, build, and maintain high-fidelity, scalable simulation platforms for Software-in-the-Loop (SIL), Hardware-in-the-Loop (HIL), and Vehicle-in-the-Loop (VIL) testing and learning.
• Realistic Virtual Environments and Agent Modeling: Enhance physics-based sensor models by combining them with deep learning and generative AI techniques. Generate synthetic data with realistic noise and distortions to test the robustness of the perception system under various conditions. Implement surrounding agents (vehicles, pedestrians) that exhibit unpredictable, human-like interactions using reinforcement learning or data-driven behavior models to train and evaluate the modular stack's social compliance and responsiveness.
• Performance Analysis and Automated Feedback Loops: Define Key Performance Indicators (KPIs) and metrics to quantitatively evaluate system performance, safety, and ride comfort. Build reporting dashboards and data pipelines that automatically feed analysis results back for algorithm improvement.
• Design and Build Simulation-Based Closed-Loop Learning Pipelines: Lead the development of large-scale simulation environments to train E2E driving policies from scratch and strengthen them through iterative trial and error. This is not just about replaying scenarios but creating a training ground that pushes the model's performance to its limits.
• Provide Reinforcement Learning (RL) and Imitation Learning (IL) Environments: Collaborate closely with Planning and Control engineers to define and implement effective reward functions, observation spaces, and action spaces within the simulation for learning. Develop curriculum learning strategies that allow the agent to progressively learn more difficult tasks.
• Sim2Real and Data-Driven Improvement: Research and apply strategies to minimize the Sim2Real gap, ensuring that models trained in simulation perform well in real vehicles. Build data pipelines that analyze simulation results (successes, failures, key metrics) to automatically feed back into the training dataset and continuously improve the model.
• Intelligent Edge Case Generation and Training: Use advanced techniques like importance sampling, reinforcement learning, or LLM-based generation to efficiently discover model failure modes. Generate edge cases and adversarial scenarios and integrate them into the training data to enhance the model's robustness.
• Collaborate with cross-functional teams, including machine learning engineers, software integration engineers, hardware platform engineers, and quality assurance, to integrate multi-vision E2E algorithms into ADAS systems.
• Participate in code reviews and knowledge-sharing sessions to foster a collaborative work environment.
• Mentor and provide technical guidance to junior/entry engineers.

형태

정규직

Application Review > Screening Test > Interview(s) > Reference Check > Offer > Onboarding