David Breeden Email & Phone Number

Mountain View, California, Us

Planner Eval • Behavior Signals • Interpretable Metrics and Contexts

Pittsburgh, Pennsylvania, Us

Technical Lead Manager, Motion Planning & Controls MetricsThe following are summaries of the highest-impact deliverables that my team produced.- Ride quality tools and their adoption. These tools provide insight that guides autonomy feature development, release promotion, and operational advisories. In collaboration with Resim Integration, Virtual Testing, and Autonomy teams, we drove an 8x increase in re-simulation testing.- Re-simulation tests, infrastructure, and tools. My team unified simulation and re-simulation test frameworks. Failures in re-simulation tests blocked release.- Team vision and growth. I grew the team from zero to seven members, including myself. I developed our team charter and yearly hiring plan.- Metrics for: ride quality and public perception, the Design Verification Plan, pickup & dropoff, remote guidance, emergency braking, and map deviations. My team developed metrics and dashboards to guide development in each of these areas. This includes Argo's automated metrics evaluating safety and traffic violations.- Automated fleet event mining. My team implemented an automated workflow to annotate a fleet ride quality event, validate that the event could be replicated in simulation, prioritize the event for human review, and then integrate the event into developer and regression testing.

Pittsburgh, Pennsylvania, Us

Technical Lead, Motion Planning & Controls MetricsI collaborated on the design and implementation of significant improvements to Argo's Data Analysis Framework (ADAF):- ingesting metrics to the data warehouse, streamlining that experience for metrics authors, and building the first Looker dashboards at Argo.- enabling richer testing capabilities- unlocking necessary performance improvements- emphasizing readability and ease of useI made ADAF more modular and hardened it, scalably increasing unit test coverage and improving dependency management.I designed and implemented centralized onboard event annotation and logging, and then implemented a driving direction monitor.

Pittsburgh, Pennsylvania, Us

Motion PlanningI:- stood up ADAF "on ingest" with the Data Core team, collecting metrics from all fleet logs.- implemented metrics for route completion, performance, and stopping location.- designed and implemented templating in test configuration, reducing redundancy.- established best practices for mocking C++ extension libraries in Python unit tests.- contributed to the onboard planning software.

Ivins, Ut, Us

We developed a legal research tool that provides access to cases and statutes with extensive metadata, and a tool that automatically analyzes user-uploaded documents with regards to arguments, drafting and context.I developed and evaluated transparent and explainable methods for:- Document parsing- Document deduplication- Entity resolution- Document classification- Sentence clustering- Topic clustering- Search query parsingI engineered systems throughout the data processing pipeline and user-facing product:- Source data ingestion- Text feature extraction- Manual review generation, assignment, UI and application- Search engine indexing- Public website- Application frontend- Internal tools frontendI developed and maintained infrastructure:- Server configuration- Server monitoring- Automated testing- Continuous integration- REST API- Payments, permissions and subscription management

San Mateo, Ca, Us

Involved in every stage of a project lifecycle, specifically including research, proposal writing, project management, requirements gathering, high- and low-level software design, Java/C++ implementation, testing, documentation and demonstration.Regularly wrote proposals for NASA and DoD contracts.Managed a Navy Phase I SBIR building a vision-based automated fire suppression system.Was lead developer for:- a Phase II SBIR developing an aggressive, autonomous path planner for aircraft under fire- a Phase I SBIR developing a tool for assessment of space weather causes in satellite anomalies- a Phase I SBIR integrating proprietary automated scheduling algorithms with NASA's open-source EUROPA planning and scheduling framework

I conducted research for the Stanford Artificial Intelligence Robot (STAIR) Computer Vision Group, developing software to generate synthetic datasets for object recognition training.

Bellingham, Wa, Us

I developed user and data acquisition interfaces for real-time image mosaicing software under Prof. Kenneth Salisbury. The software obtained video from a novel high-resolution confocal microscope and performed image registration and mosaicing in real-time. I used NI LabView to acquire the video stream and MFC for the user interface. The project is described in our 2008 SPIE paper, "Real-Time Image Mosaicing with a Dual-Axes Confocal Microscope."