Must-have

• production software for model training
• offline evaluation and release-comparison workflows
• learned models for performance prediction
• training, validation, and holdout strategies
• model and release metrics
• statistical rigor in experiments
• internal tools for software version comparison
• Python software engineering skills
• supervised ML models development
• statistics and experimental design
• ML tooling such as PyTorch
• large datasets using pandas and NumPy

Nice-to-have

• probabilistic forecasting, trajectory prediction
• classical ML methods
• calibration, uncertainty estimation
• building internal analytics or ML tools
• workflow orchestration or experimentation tools
• Bazel or other large-scale build systems
• autonomous driving experience