Computer Vision & Autonomy Engineer

Arxlight; Oakland (CA); On-site; USD 100,000 - 130,000; Full time

Job summary

Arxlight, a technology startup in Oakland, CA, seeks a Computer Vision & Autonomy Engineer. This role involves the development of perception algorithms for high-speed autonomous systems, focusing on deep learning, object tracking, and real-time processing. Candidates should have a Master’s or PhD in a relevant field and expertise in C++ and CUDA. Preference is given to those with experience in embedded systems and high-fidelity simulation. Join us in designing cutting-edge drones that save lives.

Qualifications

• Expert knowledge of object tracking and moving camera geometry.
• Proficiency in C++20 and CUDA for high-throughput image processing.
• Deep understanding of 3D geometry and filtering techniques.

Responsibilities

• Develop real-time Deep Learning and Classical CV algorithms.
• Implement Visual-Inertial Odometry and filtering techniques.
• Create GPS-denied navigation solutions for autonomy.

Skills

Object tracking

Dynamic Vision

C++20

CUDA

Python

3D geometry

Kalman Filtering

Education

Master’s or PhD in Robotics, Computer Science, or Aerospace Engineering

Tools

NVIDIA Jetson Orin

NVIDIA Isaac Sim

Unreal Engine 5

Job description

Company Overview

We are a rapidly growing technology startup focused on delivering next-generation drones for security and safety applications. Our company vertically integrates hardware and software to create leading edge capabilities in the UAV space, with a focus on saving lives.

As a Computer Vision & Autonomy Engineer, you will be joining the team responsible for the design, development, and implementation of high-speed perception and autonomy stacks capable of identifying and tracking highly dynamic objects. You will solve the unique challenges of high-dynamic sensing, where relative velocities are extreme and the margin for error is zero.

Key Responsibilities

• Perception Pipeline Development: Develop robust real-time Deep Learning and Classical CV algorithms for classification, and tracking (e.g., YOLO, Transformer-based architectures) of highly dynamic objects.
• High-Speed State Estimation: Implement Visual-Inertial Odometry (VIO) and filtering techniques to estimate target 3D trajectories and "Time-to-Go" under high-G maneuvers.
• GPS denied perception stack: Create "GPS-denied" navigation solutions and anti-jamming vision pipelines that maintain autonomy when external signals are compromised.
• Guidance Logic: Design "Vision-Based Pursuit" laws and Proportional Navigation (PN) enhancements that translate visual target states into actionable steering commands.
• Real-time Deployment: Optimize algorithms for ultra-low latency execution on low-power devices, ensuring the "sensor-to-actuator" delay is minimized.
• Deterministic Benchmarking: Profile and eliminate "long-tail" latency spikes in the autonomy stack to ensure a deterministic sensor-to-actuator response time.

Required Qualifications

• Education: Master’s or PhD in Robotics, Computer Science, or Aerospace Engineering with a focus on Computer Vision or Autonomous Systems.
• Dynamic Vision skills: Expert knowledge of object tracking (KCF, SORT, DeepSORT) and the geometry of moving camera platforms.
• Real-Time Software: Proficiency in C++20 and CUDA for high-throughput image processing, and Python for training ML models.
• Mathematics: Deep understanding of 3D geometry, Kalman Filtering (EKF/UKF), and the physics of relative motion.

Preferred Skills

• EO/IR camera: Experience working with Long-Wave Infrared (LWIR) or Mid-Wave Infrared (MWIR) sensors.
• Embedded Systems: Experience deploying models on NVIDIA Jetson Orin or FPGA-based vision processing.
• High-Fidelity Simulation: Proficiency in NVIDIA Isaac Sim, Unreal Engine 5, or Gazebo to generate synthetic data for rare "corner-case" scenarios.

Control Integration: Understanding of how perception latency affects the stability of flight control loops.