← Waymo Interview InsightsMachine Learning Engineer

Waymo·Machine Learning Engineer·Technical Phone Screen·Senior

Senior
May 2026

Summary

Waymo ML Engineer interview with a heavy focus on model efficiency and training techniques. The questions leaned pretty deep into the ML fundamentals side, stuff like QAT and contrastive learning that you don't always see covered in generic prep guides. Felt like they actually wanted to see if you'd worked with these things hands-on, not just read about them.

Questions Asked (4)

Q1

How does Quantization-Aware Training work, and what are the trade-offs compared to post-training quantization?

Technical Trade-offsSystem Design
Author's notes

This is where I spent most of my mental energy.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly defining Quantization-Aware Training (QAT) and post-training quantization (PTQ), then compare their trade-offs in terms of accuracy, performance, and deployment complexity. Emphasize how QAT simulates quantization during training to maintain accuracy, while PTQ is simpler but may sacrifice accuracy. Conclude with practical considerations for choosing between them, especially in safety-critical systems like autonomous driving.

Pro tip: Highlight that QAT is particularly valuable for low-bit quantization (e.g., 4-bit) where PTQ often fails, but note that QAT requires significant training resources and access to the original training pipeline. This shows you understand the practical constraints in real-world deployments.

1. Define QAT and PTQ

Briefly explain that QAT inserts fake quantization nodes during training to simulate quantization effects, while PTQ quantizes a pre-trained model without retraining.

2. Explain QAT mechanics

Describe how QAT works: forward pass uses quantized weights/activations, backward pass uses straight-through estimator (STE) to propagate gradients, and the model learns to be robust to quantization noise.

3. Compare trade-offs

Discuss accuracy: QAT typically achieves higher accuracy, especially at low bit-widths; PTQ is faster and easier but may degrade accuracy. Also compare resource requirements: QAT needs training data, compute, and time; PTQ is post-hoc and requires only calibration data.

4. Relate to deployment context

Tie the trade-offs to real-world scenarios: for latency-critical, resource-constrained edge devices (like in autonomous driving), QAT may be necessary for aggressive quantization, while PTQ suffices for server-side inference with higher bit-widths.

5. Conclude with recommendation

Summarize that the choice depends on accuracy requirements, available resources, and deployment constraints, and mention that hybrid approaches exist (e.g., QAT for some layers, PTQ for others).

Key Points to Mention

  • Quantization reduces model size and latency by using lower-precision data types (e.g., INT8).
  • QAT simulates quantization during training, allowing the model to adapt and often preserving accuracy better than PTQ.
  • PTQ is simpler and faster but can suffer significant accuracy loss, especially at very low bit-widths (e.g., 4-bit).
  • QAT requires access to the training pipeline, labeled data, and additional compute for retraining or fine-tuning.
  • Trade-offs include accuracy vs. resource cost, deployment complexity, and hardware support for quantized operations.
  • In safety-critical applications like autonomous driving, QAT may be preferred to ensure reliability under aggressive quantization.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q2

Walk me through how knowledge distillation works and when you'd choose it over other model compression techniques.

Technical Trade-offsAlgorithms & Data Structures
Author's notes

Felt okay here.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by clearly explaining the core idea of knowledge distillation: training a smaller student model to mimic the outputs (soft labels) of a larger teacher model. Then discuss when to choose it over other compression techniques, emphasizing trade-offs in accuracy, latency, and resource constraints, and relate it to Waymo's need for efficient real-time models.

Pro tip: Mention that distillation can be combined with other techniques like quantization or pruning for compounded benefits, and highlight that the choice depends on whether you have a high-capacity teacher and whether the task benefits from soft targets (e.g., ambiguous classes).

1. Explain the core mechanism

Describe how knowledge distillation works: a teacher model's softmax outputs (soft labels) are used as targets for a student model, often with a temperature parameter to soften probabilities. Mention the loss function combining distillation loss and student's own cross-entropy.

2. Compare with other compression techniques

Contrast distillation with pruning, quantization, and low-rank factorization. Highlight that distillation transfers knowledge to a new architecture, while others modify an existing model. Note that distillation can be used alongside them.

3. Discuss when to choose distillation

Explain scenarios: when you have a large, accurate teacher and need a smaller, faster model; when the task has soft inter-class relationships (e.g., similar classes); when you want to ensemble multiple teachers into one student; or when you need to deploy on edge devices with limited compute.

4. Address trade-offs and Waymo context

Mention potential downsides: requires training a teacher (costly), may not work if teacher is too complex for student capacity, and can be sensitive to hyperparameters. Relate to Waymo: real-time perception models need low latency and high accuracy; distillation can help compress large models for onboard deployment.

Key Points to Mention

  • Soft labels vs. hard labels: soft labels provide more information about class similarities.
  • Temperature parameter: controls the softness of probability distributions.
  • Loss function: typically a weighted sum of distillation loss (KL divergence) and student loss (cross-entropy).
  • Comparison with pruning and quantization: distillation changes model architecture, while pruning/quantization modify existing weights.
  • Use cases: model compression for deployment, ensemble distillation, and transfer learning.
  • Waymo relevance: need for efficient, low-latency models for autonomous driving perception and prediction.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q3

What metrics would you use to evaluate a machine learning model, and how do you decide which ones matter for a given problem?

Product Analytics & MetricsA/B Testing & Experimentation
Author's notes

Blanked for a second on how to scope this without knowing the specific task.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by acknowledging that metric choice depends on the problem type, business goals, and data characteristics. Then walk through a structured decision process: identify the task (classification, regression, ranking, etc.), consider class balance and error costs, and select metrics that align with stakeholder needs. Emphasize that no single metric is universally best; you must weigh trade-offs and validate with real-world impact.

Pro tip: For safety-critical systems like autonomous driving, always prioritize metrics that capture rare but catastrophic errors (e.g., false negatives in pedestrian detection) and consider using precision-recall curves over ROC when classes are highly imbalanced.

1. Clarify the problem and business objective

Understand the task type (classification, regression, ranking, etc.) and what the model's output will be used for. Identify the cost of different error types (false positives vs. false negatives) and any regulatory or safety constraints.

2. Consider data characteristics

Assess class balance, dataset size, and whether the data is time-series or spatial. This influences whether metrics like accuracy, F1, AUC-ROC, or precision-recall are appropriate.

3. Select candidate metrics

Choose metrics that align with the objective: for classification, consider accuracy, precision, recall, F1, AUC-ROC, AUC-PR; for regression, MSE, MAE, R²; for ranking, NDCG, MAP. Also consider calibration metrics if probabilities matter.

4. Evaluate trade-offs and align with stakeholders

Discuss how metrics trade off (e.g., precision vs. recall) and which aligns with business KPIs. Use a validation set or cross-validation to compare. Consider composite metrics or business-specific metrics (e.g., cost savings).

5. Validate with online experiments and monitor

Ultimately, offline metrics are proxies. Propose A/B testing or shadow mode to measure real-world impact. Set up monitoring for drift and ensure metrics remain relevant over time.

Key Points to Mention

  • Task type and business objective drive metric selection (e.g., safety-critical vs. user engagement).
  • Class imbalance: prefer precision-recall AUC over ROC AUC when positives are rare.
  • Cost-sensitive evaluation: assign costs to FP/FN and use expected cost as a metric.
  • Calibration: if predicted probabilities are used for decisions, check calibration (e.g., reliability diagrams).
  • Online vs. offline metrics: offline metrics are proxies; validate with A/B tests or shadow deployment.
  • Domain-specific metrics: for Waymo, consider disengagement rate, collision rate, or near-miss metrics.

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.

Q4

Explain contrastive learning and describe a scenario where you'd apply it.

Algorithms & Data StructuresTechnical Trade-offs
Author's notes

Actually enjoyed this one.

Create a free account to read the full note

AI HintsAI Generated

Suggested Approach

Start by defining contrastive learning clearly, emphasizing its goal of learning representations by pulling similar samples together and pushing dissimilar ones apart. Then, describe a concrete scenario relevant to Waymo, such as learning robust representations from unlabeled sensor data for object detection or retrieval, and explain how you would implement it, including the choice of positive/negative pairs and loss function.

Pro tip: Connect contrastive learning to Waymo's specific challenges, like handling multimodal sensor data or improving few-shot learning for rare objects, and discuss trade-offs such as computational cost and the need for large batch sizes.

1. Define Contrastive Learning

Explain that it's a self-supervised technique that learns representations by contrasting positive pairs (similar) against negative pairs (dissimilar) in an embedding space.

2. Explain the Mechanism

Describe how it works: define an anchor, positive, and negative samples; use a loss like InfoNCE or triplet loss to minimize distance between anchor and positive and maximize distance from negatives.

3. Describe a Relevant Scenario

Choose a scenario at Waymo, e.g., pre-training on unlabeled LiDAR or camera data to learn features for downstream tasks like object detection or segmentation, especially when labeled data is scarce.

4. Outline Implementation Details

Discuss how you would generate positive/negative pairs (e.g., augmentations, temporal consistency), select a loss function, and handle challenges like large batch sizes or memory banks.

5. Highlight Trade-offs and Benefits

Mention benefits like improved generalization and label efficiency, and trade-offs such as computational cost and the need for careful negative sampling.

Key Points to Mention

  • Self-supervised learning paradigm
  • Positive and negative pair construction
  • Contrastive loss functions (e.g., InfoNCE, triplet loss)
  • Application to unlabeled sensor data (LiDAR, camera) for pre-training
  • Handling large batch sizes or memory banks
  • Trade-offs: computational cost vs. representation quality

AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.