Deep Learning Engineer
VERDANT ROBOTICS
Deep Learning Engineer Job Description
Hayward, CA - Full-Time Position
Verdant Robotics aims to lead the second agricultural revolution with our AI-powered computer vision platform. Deep learning is a core part of our onboard software, both retraining existing models on our data and developing new application-specific models to solve unique problems on the farm. We are looking for a deep learning engineer with experience or an interest in field robotics to drive the development of our deep learning capabilities.
As a Deep Learning Engineer, you will develop models using TensorFlow, deploy them in a C/C++ environment on board the robot, develop tools for analyzing performance both during training and in the field, test and debug your code on the robot, and make improvements to our data and training pipelines. We’re open to candidates of varying experience levels, but working knowledge of C/C++, Python, and Tensorflow is required. Most importantly, you should be driven to make an impact with your work by rapidly reaching the market with a new robotic system doing useful work for paying customers. Most of us at Verdant have worked at self-driving car companies (Google/Waymo, Zoox, Cruise) and realize that we can make an impact and a profitable business on a much shorter timescale with field robotics.
Our core technology empowers farmers to grow more profitable food. We use computer vision and artificial intelligence to understand every part of every plant at a massive scale -- and with automated robotics, we close the loop and act on that information in real-time. We are leaders in robotics, computer-vision, and machine learning. Together with our partners, we are building environmentally-friendly, sustainable and highly productive farms.
Our office is located in Hayward, CA about 25-30 mins average commute from most places in the Bay Area. We offer competitive pay and a great team environment.
The Role: Deep Learning Engineer
What You Will Do:
- Train off-the-shelf TensorFlow models (e.g. RetinaNet) for detection and classification in camera images
- Deploy trained models on the robot and integrate them with the rest of the stack
- Optimize models for runtime, including use of TensorRT
- Adapt existing models, or develop new models, for unique vision capabilities on the farm
- Develop metrics for model performance during the training and in the field
- Debug failures and identify root causes throughout the stack--whether related to timing, image quality, programming errors, or incorrect assumptions about the environment
- Help identify architectural changes to improve on-robot performance (e.g. multi-frame models, changes to image resolution or frame rate)
- Help select a scalable training platform (e.g. Cloud TPU versus multi-GPU workstations)
- Identify and implement pipeline changes to improve data labeling and cleaning
- Regularly test your code on the robot to understand system-level performance
Minimum Qualifications
- BS in CS, EE, or related field
- Experience developing image-based models in TensorFlow (including hyperparameter tuning, data curation, detecting train/test/eval skew, exporting graphs for inference in C++ runtime environment)
- Coursework or experience in traditional computer vision
- Fluent in C++ and Python
Preferred Qualifications
- MS or PhD in CS, EE, or Robotics; or several years of experience with complex robotics systems operating outside the lab
- Graph optimization with TensorRT
- CUDA programming
- Some exposure to realtime / embedded programming
Email Resume to info@verdantrobotics.com
Verdant Robotics provides equal employment opportunities (EEO) to all employees and applicants for employment without regard to race, color, religion, sex, national origin, age, disability, genetics, sexual orientation, gender identity or gender expression.