Perception Engineer

Who We Are:

Aurelius Systems is a VC backed defense tech startup building autonomous, edge deployed directed energy systems for counter-UAS. We build laser weapons to shoot down drones.

We're a small team of engineers, former US military operators, and subject matter experts scaling America's directed energy dominance. The first cost effective, reliable and robust laser weapon system.

Our namesake isn't an accident. Marcus Aurelius wrote about doing the work in front of you, every day, without excuses. Henry Ford didn't wait for permission to reinvent manufacturing. That's how we operate - small team, unreasonable output, no hiding behind the unachievable.

In addition to our San Francisco lab, we’ve opened a Detroit manufacturing hub and we field test weekly on our own 400-acre private range.

The Role and Your Impact:

We are seeking a skilled Perception Engineer to join our software team and work on end-to-end perception and sensor-fusion stack. You will develop, train, and deploy vision and sensor based models; manage data pipelines; and enable real-time detection and tracking for our laser-based defense system. This role bridges data science, software engineering, and robotics to deliver reliable, high-throughput perception performance on edge hardware.

What You'll Own:

• Design, train, validate and fine-tune machine-learning and deep-learning models (e.g., YOLO, RT-DETR, CNNs) for object detection, classification, and segmentation.

• Integrate and fuse data from multi-modal sensors (RGB, thermal, LiDAR/ToF, IMU, encoders) to produce robust, real-time Regions of Interest (ROIs).

• Research, implement, and as-needed develop high and low-level image-processing techniques, such as deconvolution, low SNR detection, and motion-isolation techniques.

• Collaborate with hardware teams to integrate and troubleshoot sensors (global-shutter and rolling-shutter cameras, thermal imagers, LiDAR/ToF modules, IMUs) over GigE Vision, USB3 Vision, CAN, SPI, and I²C protocols; develop and debug embedded firmware in C/C++ (or Rust) for microcontrollers (STM32, NXP, TI) and FPGAs using VHDL/Verilog within RTOS environments (FreeRTOS, Zephyr).

• Build scalable data ingestion, labeling, augmentation, and storage pipelines (simulated and field data) ensuring 100k+ labeled frames accuracy.

• Optimize inference frameworks for edge deployment (GPU/FPGA), achieving ≥500 Hz end-to-end throughput.

• Develop dashboards and telemetry for drift analysis, hardware health monitoring, performance metrics, and automated retraining triggers.

• Author clear technical docs; mentor junior engineers on best practices in vision, sensor-fusion, and embedded firmware engineering.

• Determines development needs by directly analyzing technical and physical limitations of our goals.

What We're Looking For:

• 2-6+ years in computer vision, sensor fusion, or robotics perception roles

• Strong C++ with deep ML engineering experience

• Hands-on with ML frameworks (TensorFlow, PyTorch) and real-time inference engines (TensorRT, OpenVINO)

• Computer vision, tracking, and detection in real-time, real-world conditions

• Familiarity with ROS2, Docker, and CI/CD for ML pipelines

• Experience with multi-sensor calibration and data synchronization

Where you probably come from: Perception roles at defense companies, autonomous vehicle programs, robotics platforms, or aerospace programs that deploy perception against real targets.

We want to talk if: You've shipped a perception system that ran on real hardware against real targets. You know the gap between paper SOTA and what survives field conditions.

Not a fit if: Your background is research only without deployment, or your CV experience is offline batch processing only.

Nice to Haves:

• Edge-AI optimization (quantization, pruning)

• Experience with FPGA or embedded GPU platforms

• Background in defense or safety-critical systems

• Familiarity with cybersecurity guidelines and secure coding practices

Education:

• BS, MS, or PhD in CS, EE, Robotics, or equivalent. Track record matters more than degree.

How You Operate:

• Extreme bias for action. You ship working perception on real hardware, not slideware

• You debug from first principles, not intuition alone

• Comfortable with ambiguity and fast iteration in a startup environment

• Clear communicator across software, hardware, and operator-facing surfaces

• Self directed. You identify what needs to happen next and do it without being told

Why Join Aurelius Systems?

• Build more in 1 month than most engineers build in 1 year. We field test weekly. Your work goes downrange, not into a filing cabinet.

• Career velocity is real. At ~10 engineers, there are no layers between you and impact.

• Work on a problem that actually matters. Small cheap drones are changing warfare. Our laser systems are the asymmetric answer - infinite magazine, near zero cost per shot, scalable to every base, border, facility, and truck.

• Join the densest defense startup ecosystem in the country. California is where the next generation of defense companies are being built.

How We Work:

Core hours are Monday through Friday, 9 to 6. When we're sprinting toward a demo or field test, the team ramps up - nights, weekends, whatever it takes to ship. When the sprint lands, we ramp down. We don't manufacture intensity for show.

Benefits:

• Competitive salary + equity

• United Health Care medical, dental, and vision coverage

• Flexible 18 days PTO + 5 sick days

• Travel to field test events and range days

• Covered daily lunches and office snacks + drinks

• E-bike / scooter stipend ( Up to $500)

• Direct access to leadership and real ownership over your work

Export Control Notice:

This role requires access to export-controlled information or items that require "U.S. Person" status. As defined by U.S. law, individuals who are any one of the following are considered to be a "U.S. Person": (1) U.S. citizens, (2) legal permanent residents (a.k.a. green card holders), and (3) certain protected classes of asylees and refugees, as defined in 8 U.S.C. 1324b(a)(3).