At May Mobility, a Machine Learning Engineer plays a pivotal role in defining the future of autonomous transit. Unlike traditional automotive companies focusing solely on personal passenger vehicles, May Mobility designs and deploys autonomous micro-shuttles that integrate directly into municipal transit systems. This unique mission requires highly reliable, safety-critical machine learning systems capable of navigating complex urban environments. Your work will directly impact the safety, efficiency, and scalability of these autonomous fleets.

As part of the engineering team, you will tackle some of the most challenging problems in spatial computing, deep learning, and computer vision. Machine learning is embedded across the entire autonomy stack, but it is particularly critical in perception, prediction, and mapping. For those joining the mapping and routing teams, the focus is on automating the generation of high-definition (HD) maps, understanding lane-level connectivity, and building robust route networks. You will transform raw sensor data, including LiDAR, camera, and GPS trajectories, into highly structured, semantically rich graph networks that the vehicle's planner can query in real time.

This role is intellectually demanding and highly cross-functional. You will collaborate closely with software engineers, infrastructure specialists, and product managers to move models from research to production. At May Mobility, engineering is driven by a safety-first culture where code quality, deterministic fallbacks, and rigorous validation are just as important as model accuracy. If you are passionate about applying state-of-the-art deep learning to real-world physical systems, this role offers an unparalleled opportunity to see your models driving on public roads daily.

To excel in the Machine Learning Engineer interview loop at May Mobility, you must prepare for a blend of core software engineering, deep learning theory, spatial system design, and behavioral scenarios. The following questions are representative of the patterns observed in real interview loops for this position.

Machine Learning & Deep Learning Theory

These questions assess your foundational understanding of ML algorithms, optimization techniques, and their application to spatial or sequential data.

• Explain how Graph Neural Networks (GNNs) can be used to predict lane connectivity or road topology from raw sensor outputs.
• What are the trade-offs between using a Transformer-based architecture versus a Convolutional Neural Network (CNN) for processing sequential spatial data like vehicle trajectories?

Preparing for an interview at May Mobility requires a structured approach that balances deep technical expertise with an understanding of physical, safety-critical systems. You should focus your preparation on demonstrating both your theoretical depth and your practical engineering execution.

Role-Related Knowledge – You must demonstrate a deep understanding of modern ML techniques, particularly those applicable to spatial data, computer vision, and graph representation learning. Be prepared to explain not just how an algorithm works, but why you chose it, its limitations, and how it behaves under edge-case conditions.

System Design & Scalability – Autonomous vehicle systems require highly optimized pipelines, both on the vehicle and in the cloud. You need to show that you can design robust, scalable architectures that process large datasets efficiently, maintain low latency during inference, and feature deterministic safety fallbacks.

Software Engineering Rigor – Excellent modeling is useless without high-quality code. You will be evaluated on your ability to write clean, modular, and well-tested code in Python or C++. Pay close attention to algorithmic complexity, memory management, and edge-case handling during coding sessions.

Collaboration & Safety Culture – At May Mobility, safety is paramount. You must show that you prioritize rigorous validation, clear documentation, and cross-functional collaboration. Showing an awareness of how your models impact downstream planning and control modules is key to demonstrating strong systems-level thinking.

The interview process at May Mobility is designed to evaluate your technical capabilities, system-level thinking, and cultural alignment through a series of structured conversations and practical assessments. The loop is highly collaborative, giving you the opportunity to interact with multiple members of the engineering and leadership teams.

The process typically begins with a conversational screen with a recruiter to align on your background, career goals, and compensation expectations. This is followed by a technical phone screen, which usually consists of a coding assessment focused on data structures and algorithms (often with a spatial or geometric twist) or a deep-dive discussion on your machine learning background. Successful candidates then progress to the virtual onsite loop.

The onsite loop is comprehensive, consisting of several distinct rounds. You will face a machine learning system design interview, a deep-dive session on your past technical projects, a coding and hands-on software engineering round, and behavioral interviews with engineering leaders. Throughout the loop, interviewers will assess your ability to solve ambiguous problems, communicate complex technical trade-offs, and work effectively within a safety-critical engineering environment.

The visual timeline above outlines the typical progression of the May Mobility interview loop, from the initial touchpoint to the final decision. Candidates should expect the entire process to take between three to five weeks, depending on scheduling availability. Use this timeline to pace your preparation, ensuring you allocate ample time for both coding practice and system design review before the onsite loop.

To pass the technical bar at May Mobility, you must demonstrate mastery across several core domains. Below is a detailed breakdown of the primary evaluation areas you will encounter during the interview loop.

Mapping & Spatial Machine Learning

This evaluation area focuses on your ability to apply deep learning to spatial data, vector maps, and geometric networks. Since mapping is foundational to May Mobility's autonomy stack, you must show how you can automate the parsing of physical environments into structured digital representations.

Be ready to go over:

• Graph Neural Networks (GNNs) – Representing road networks as graphs, node/edge classification, and predicting connectivity.
• Sensor Fusion for Mapping – Combining sparse LiDAR point clouds, GPS trajectories, and camera imagery to generate semantic map layers.
• Geometric Deep Learning – Processing non-Euclidean data structures and ensuring rotational/translational invariance in spatial models.
• Advanced concepts (less common) – Vectorized map auto-regressive generation, lane centerline extraction using transformer-based decoders, and topological loss formulations.

Example questions or scenarios:

• "How would you design a neural network that takes a raw, noisy point cloud of an intersection and outputs a vectorized graph of the drivable lanes?"
• "Explain how you would use temporal GPS trajectories from fleet vehicles to identify when a physical road layout has changed, triggering an automated map update."

ML System Design & Scalability

Building models that run in research environments is different from deploying them to run reliably on physical vehicles or in massive cloud-based map-generation pipelines. This area tests your ability to design production-grade ML systems.

Be ready to go over:

• Data Pipeline Architecture – Designing efficient ETL pipelines for processing multi-modal sensor logs.
• Model Optimization & Inference – Quantization, pruning, and optimizing models for deployment on edge hardware (e.g., NVIDIA Jetson or TensorRT).
• Validation & Testing Frameworks – Designing offline simulation tests, shadow deployments, and regression testing pipelines.
• Advanced concepts (less common) – Active learning loops for automatic edge-case detection, distributed training strategies for large spatial-temporal models.

Example questions or scenarios:

• "Design a scalable cloud infrastructure that automatically ingests daily driving logs, extracts instances of poor localization, and uses them to retrain our mapping models."
• "How would you design a multi-task learning model that shares a backbone network to predict both lane boundaries and traffic sign locations, ensuring it meets strict real-time latency requirements on the vehicle?"

Core Software Engineering & Geometry

This area evaluates your programming fundamentals and your ability to translate mathematical and geometric concepts into clean, efficient, and maintainable code.

Be ready to go over:

• Spatial Indexing & Querying – Implementing and utilizing spatial data structures like R-trees, quadtrees, or k-d trees.
• Computational Geometry – Algorithms for polygon intersection, point-in-polygon testing, and line segment simplification (e.g., Douglas-Peucker).
• Algorithm Design & Complexity – Writing highly optimized code in Python or C++, with a strong focus on time and space complexity.
• Advanced concepts (less common) – Graph search algorithms (A*, Dijkstra) optimized for complex, multi-layered road networks.

Example questions or scenarios:

• "Write a function that takes a set of overlapping bounding boxes and merges them into a single, minimal bounding polygon."
• "Explain how you would optimize a spatial query that needs to find all traffic lights within 50 meters of an autonomous shuttle moving at 15 meters per second."