AI SW Engineer

Role Overview

The WoRV team builds Mobile Manipulation products that solve driving + attachment control together in brownfield environments.

This role converts the logs, video, sensor data, operational events, and failure cases that robots leave behind in the field into data assets and evaluable evidence that can be used to improve AI products.

We work across Simulation, Hardware, and the AI autonomy stack (VLA, Motion Planning, Localization, Control). Some capabilities are implemented as explicit modules; others are extended as learned policies or VLA models. What matters is not just collecting more data, but transforming real-world customer problems into reproducible datasets, evaluation frameworks, and training/validation loops.

The AI SW Engineer designs and implements the AI improvement loop — from data collection to curation, storage/retrieval, training/evaluation, failure analysis, and back to re-collection. The goal is not a one-off pipeline for a single project, but Safe, Reliable, and Scalable product capabilities that can be applied repeatedly across multiple customers and field sites.

Responsibilities

• Design and develop data pipelines for robot and Mobile Manipulation systems.
• Build teleoperation systems for robots and Mobile Manipulation, contributing to model performance improvement through intuitive user interfaces.
• Build systems that collect, curate, store, and search field logs, video, sensor data, rosbag, control/driving events, and operational results.
• Design dataset generation, metadata management, data quality management, and labeling/curation workflows.
• Implement data interfaces needed for model training, VLA/policy evaluation, and control improvements.
• Collaborate with PM / AI / Control / Simulation teams to convert customer pain points into data-driven improvement tasks.
• Improve systems considering data pipeline latency, cost, reliability, and maintainability trade-offs.

Qualifications

• Experience implementing data processing, automation, and backend systems in Python
• Experience building at least one data pipeline: ETL, batch processing, streaming, or workflow automation
• Understanding of SQL/NoSQL databases, object storage, file formats, indexing, and metadata design
• Experience working with at least one of: logs, video, sensor data, time-series data, or unstructured data
• Interest in structuring data not as mere storage, but as assets usable for analysis, training, evaluation, and decision-making
• Ability to develop, debug, deploy, and automate operations in Linux environments
• Ability to reason about trade-offs beyond accuracy: latency, cost, reliability, and maintainability
• Ability to quickly learn new domains and data formats, and pivot to simpler approaches when needed

Preferred Qualifications

• Experience building ML training datasets, data versioning, experiment tracking, and reproducible experiment environments
• Experience building model training/evaluation pipelines, benchmarks, evaluation harnesses, and regression test datasets
• MLOps or DataOps experience
• Experience with robotics data formats: ROS2, rosbag, RGB/Depth camera, IMU, LiDAR
• Experience with large-scale data processing, high-performance I/O, parallel processing, caching, indexing, and search optimization
• Experience with brownfield environments, noisy sensors, partial observability, and edge case data problems
• Experience building data labeling tools, annotation workflows, QA pipelines, or human-in-the-loop data systems
• Experience operating data systems with Docker, CI/CD, workflow schedulers, or cloud/object storage

Who We're Looking For

• Someone who asks which data actually improves the product, rather than just collecting more
• Someone who understands the difference between model performance and product performance, and designs data pipelines as part of the AI improvement loop
• Someone who can structurally analyze failure causes from logs, video, sensor data, and operational results
• Someone with a strength in organizing messy, complex field data into forms the team can use for decision-making
• Someone who wants to build reusable data capabilities across multiple customers and projects, not one-off analyses or ad-hoc pipelines
• Someone who can design systems considering both immediate customer value and future scalability
• Someone who learns new technologies quickly but is willing to simplify boldly when a simpler approach is the right one