David Soberanis - Portfolio

Experience

12/2023 – Present

Architected and engineered the entire software stack for industrial silicon wafer feed and test machines, capable of processing 1,066 wafers per day, saving $100k per year in labor costs.
Developed embedded state machines using Python and C++ on Nvidia Jetson Nanos, handling multiple inputs/outputs and Modbus 232 communication.
Implemented a RESTful API using Flask, enabling real-time communication between machines and integration with external test equipment.
Created a responsive front-end user interface using jQuery, HTML, and CSS for real-time system control and status updates.
Designed and implemented the electronic controls circuit, integrating various sensors, motors, and safety features.
Engineered critical safety features, including an E-stop system and laser curtain, passing TUV electrical inspection standards.
Implemented an efficient barcode scanning system for wafer tracking and identification.
Developed a comprehensive logging and diagnostic system using Python's logging module and Tailscale for remote troubleshooting and performance analysis.

Programmed an embedded state machine in C++ on an Arduino Nano to control a complex system of solenoids and sensors, processing over 3,840 locks and latches per day.
Optimized system performance to achieve cycle times of less than 5 seconds per lock/latch placement operation.

Engineered an embedded control system using C++ on Arduino Nano to manage multiple screw feeders, pneumatic solenoids, and reed sensors, processing up to 7,680 screws per day.
Developed a custom communication protocol between Arduino Nano and Jetson AGX Xavier, featuring asynchronous messaging and real-time state reporting.

Programmed in C++ to develop firmware for an Arduino Mega, integrating Bluetooth technology with serial communication for control of blades and movement with a smartphone. Total cost was reduced by a factor of 8.
Engineered an automated lawn mower controlled via smartphone that does not require an underground perimeter wire, achieving 86% reduction in hardware costs per unit from industry standard of $4000 to $580.
Implemented advanced control systems, including an H bridge to activate the linear actuator to engage blade control and a relay for engine start, then Hall effect speed controllers to drive the Hub Motors for seamless movement.
Awarded $1000 from UCSD's The Basement incubator, demonstrating the project's potential to secure funding.

Designed, programmed, and built Computer Vision and Reinforcement Learning trained autonomous scale vehicles under Dr. Silberman at UCSD.
Achieved multiple wins at AI scale car races:
- 1st place at 2019 Balboa San Diego wide race
- 1st place at 2019 UCSD race
- 6th place at Oakland Statewide race
Participated in 3 San Diego races and 4 California state races.
Leveraged OpenCV (Open Computer Vision) and DonkeyCar (based on Keras) and Nvidia Jetson Nano at Triton AI to drive cutting-edge development and performance.