Elton Shon

Founding engineer (employee #2) Backed by YC, Khosla Ventures, Lenovo, Miracle Plus, Xiaomi, and AlphaX, Industrial Next is building the future of manufacturing through automated lines driven by end-to-end robot learning. - Architected firmware and software stack for sensors, vision systems, and industrial robots, enabling seamless integration across hardware and AI. - Co-developed data collection infrastructure and foundation models for industrial manipulators, powering scalable end-to-end robot learning systems.

Dojo (Autopilot Hardware), Firmware and Software

- Dojo - Implementing an image-based pose estimation vision system in C++ using OpenCV, Haar Cascade, TensorFlow object detection API, template matching and point registration techniques - Implementing ROS control communication between Autopilot computer and motor drives using CANopen protocol - Implemented a motion/visualization library for an industrial robotic arm in C++ using ROS MoveIt! and Rviz - Designed and 3D printed mounts for camera and skateboard motor using CATIA - Implemented a QR code finder/reader in Python using OpenCV and a Raspberry Pi camera - Generally supported Model 3 ramp

• Designed and implemented embedded firmware in C to monitor and control an automated temperature control cell test system using PID and communication protocols such as SPI and CAN • Implemented a data pipeline and a standalone executable GUI in Python to control and gain feedback from the automated temperature control system using Python libraries such as tkinter, threading, pySerial, and adodbapi • Developed test tools in Python to validate the temperature control system using pySerial and matplotlib libraries • Streamlined cell test fixtures installation process and led retrofits for testing channels at the Gigafactory

- Implemented yellow tactile, stop sign, and traffic light detection and recognition on edge devices with Raspberry Pi Cameras in Python using OpenCV, Haar Cascades, and Convolution Neutral Networks. - Developed an automated data labeling pipeline by customizing You Only Look Once (YOLO) in C to detect and crop objects from images in real time, eliminating manual labeling effort.

Built a visual–kinematic data collection pipeline for robotic cars, enabling automated driving research with deep learning

- Built a Python application using OpenCV and path-search algorithms to calculate the optimal path and dispensing area for an epoxy-dispensing mechanical system, improving precision and efficiency. - Developed a data pipeline with a UI in Python/VBA that streamlined the offline wire-bonding and part pick-and-place process, reducing cycle time from 7 days to ~3 days. - Created wire-bonding diagrams for circuit boards using AutoCAD and BWX offline programming software, ensuring accuracy and manufacturability in bonding workflows.

Led interactive sessions incorporating collaborative learning activities and study strategies to deepen students’ understanding of course material.