Simulation Engineer Intern

Requirements

• Generative AI & LLM
  
  Experience:
  
  Strong understanding of Large Language Models, including prompt engineering, API integration, and structuring LLM outputs (e.g., JSON parsing)
• Programming Proficiency:
  Strong, hands‑on programming skills in Python for machine learning workflows and scripting
• Systems Thinking:
  Ability to translate abstract, real‑world concepts (like traffic rules and driving environments) into structured, programmable data and logic
• Problem Solving: A highly analytical mindset with a passion for tackling complex challenges at the intersection of AI and physical‑world simulation
• (Desirable) Simulation Software:
  Familiarity with autonomous vehicle simulation platforms (e.g., CARLA, LGSVL, or proprietary AV simulators) and procedural generation
• (Desirable) Autonomous Driving Context:
  Knowledge of autonomous driving concepts, Operational Design Domains (ODDs), ADAS systems, or NHTSA safety guidelines
• (Desirable) Formal Logic/Rules Engines:
  Exposure to formal logic, deontic logic, or building rule-based expert systems
• (Desirable) Synthetic Data Generation:
  Prior experience using generative AI for synthetic data generation, automated testing, or scenario creation

• LLM Integration:
  Leverage Large Language Models (LLMs) to architect and implement a pipeline that automatically generates realistic, highly structured driving scenarios
• Translate ODDs to Prompts:
  Convert complex Operational Design Domains (ODDs) and safety/NHTSA guidelines into effective prompt frameworks and programmatic constraints for the model
• Simulation Integration:
  Interface the LLM-generated scenarios directly with the autonomous vehicle simulation environment to create executable, varied testing grounds
• Apply Deontic Logic:
  Incorporate rule-based constraints (deontic logic) to ensure the generated scenarios strictly adhere to or properly test complex traffic laws, rights‑of‑way, and safety protocols
• Quality Assurance & Iteration:
  Collaborate with the simulation and machine learning teams to evaluate the realism, diversity, and edge‑case coverage of the synthetic scenarios, refining the models as needed