Senior Edge AI Perception Engineer

Aion Robotics Corporation•Arvada, CO

21h

About The Position

AION Robotics is seeking a skilled Edge AI Perception Engineer to design, optimize, and deploy deep learning models for real-time autonomous vehicle perception. The role focuses on building efficient inference pipelines for NVIDIA embedded platforms, including Orin and Thor, integrating perception outputs with LiDAR and camera data, and optimizing latency, throughput, and accuracy for real-world autonomous systems. The engineer will develop and optimize neural network pipelines, use NVIDIA tools and hardware resources for real-time execution, enhance perception through geometric transforms and terrain mapping, maintain data collection and model-training workflows, and implement camera pipelines on embedded platforms. The role also involves collaborating with hardware engineers on sensor calibration and synchronization, and researching and validating new perception models and methods for production use. AION Robotics builds autonomous vehicles for critical infrastructure monitoring, construction, mining, and energy applications. The company describes this role as an opportunity to contribute to next-generation perception systems operating in demanding field conditions.

Requirements

Direct real-world experience in computer vision, deep learning, or edge AI systems.
Strong proficiency with CUDA, TensorRT, NVIDIA DeepStream, TAO Toolkit, and PyTorch or TensorFlow.
Demonstrated experience with model optimization techniques such as pruning, quantization, distillation, and scheduling.
Hands-on experience deploying AI pipelines to embedded edge platforms, preferably NVIDIA Jetson Orin or Thor.
Expertise in object detection, semantic segmentation, and depth estimation.
Solid understanding of real-time embedded system constraints and low-latency optimization strategies.
Strong programming skills in C++ and Python.

Nice To Haves

Experience with low-level camera system integration, including ISP tuning, intrinsic and extrinsic calibration algorithms, and multi-camera synchronization and fusion.
Familiarity with sensor-fusion pipelines combining camera, LiDAR, and IMU.
Direct experience working in autonomous vehicles, robotics, ROS2, or safety-critical perception systems.

Responsibilities

Build and manage neural network pipelines optimized for edge deployment in autonomous vehicle systems.
Implement, compress, and optimize models using techniques such as pruning, quantization, and scheduling for GPU, DLA, and Tensor Core execution.
Work with neural network architectures including monocular depth models, YoloX, PeopleNet, and ResNet.
Create high-performance perception pipelines using NVIDIA DeepStream, TensorRT, CUDA, and TAO Toolkit.
Manage model and hardware resource allocation across GPU and DLA for real-time scheduling and execution.
Optimize end-to-end “lens-to-detections” pipelines to meet low-latency constraints on embedded devices such as NVIDIA Orin and Thor.
Apply real-time geometric transforms to object detections and semantic segmentation results for geo-referencing and LiDAR point-cloud filtering.
Develop CUDA-accelerated 3D terrain mapping.
Develop and maintain automated data collection pipelines, including dataset formatting, CVAT labeling workflows, and fine-tuning for custom training.
Implement real-time dewarped camera pipelines using GMSL drivers, VIC, and ARGUS APIs on embedded platforms.
Collaborate with hardware engineers to achieve consistent calibration and synchronization across multiple sensors.
Track advancements in neural networks, edge AI optimization, and autonomous perception, and rapidly prototype and validate new models and methods for production deployment.