Julian Kent

Leading Localization team, providing low latency and high accuracy location data for Hexagon's humanoid robot, Aeon.

Leading the team developing new features (eg. obstacle avoidance) and validating and improving control algorithms on the new Wingtra drone, Ray.

Design, architecture and implementation for 2nd Generation video ID verification system. Introduction of CI/CD and automated test hierarchies to existing products while improving codebase standards. Interviewing, hiring and mentoring of team members. Providing deep expertise and guidance on a wide variety of topics from systems engineering and Linux deployments to classical computer vision and performance tuning.

Mixed role split between project manager, technical lead and product manager. On technical side, mostly architecture design and feedback, with minor features implemented in order to better understand architecture and development bottlenecks. Product side is coordinating (and negotiating) with various stakeholders on the feature roadmap for maximum cross-customer product value while still meeting hard delivery constraints for individual customers. Managing 8 engineers.

Leading Autonomy team, responsible for developing next-generation drone capabilities via computer vision, path planning and simulation. Camera-based localization (VIO), automatic obstacle avoidance route replanning and emergency collision prevention systems, next-generation compute platform specification and development. Work performed both as integration with customer projects as well as contribution to PX4 open source ecosystem.

Hiring and managing a new team, setting up a new office, building a new agriculture-specific photogrammetry pipeline (from high-level architecture to low level implementation) and coordinating this with UI development in order to deliver features quickly. Balance of ~50% management and planning, 50% development.

Part of the team working on calibration - computing the camera poses and lens distortions via non-linear optimisation. C++11, Eigen, Ceres Projects: 360 camera and >180 degree fisheye support, internal tooling for accuracy testing, better scaling for large projects, camera and point uncertainty and parameter correlations.

Worked on a wide range of things: Software development: embedded C++ with valgrind memory analysis, embedded C, Open AT, Google App Engine (Python), PyQt Data: predictive analytics (sales predictions for stock management), geolocation (localisation algorithms for when GPS is unavailable) Electronics: analogue & digital circuit design and validation, component placement for noise & power reduction, test jig development, failure analysis as well as designing and validating fixes. Process: Agile (KANBAN) for all development (software & hardware), fully automated testing/CI/CD for both server and embedded platforms, with code, interface and full-path tests.

Performing a pilot study for geo-location systems using cell towers instead of GPS.

Lab: Robotics and Agents Research Laboratory (RARL) Dissertation entitled "Simultaneous Localisation and Mapping for Urban Search and Rescue"