软件工程师, 感知
Software Engineer, Perception
职位描述
Valkyrie Aero 正在寻找一名软件工程师，为机载任务系统开发计算机视觉和传感器融合软件。该职位的软件工程师将开发实时检测、分类和跟踪软件，将多个传感器输入融合成一个跟踪画面，并将跟踪状态传递给控制软件和操作员界面。这包括模型开发与评估、针对嵌入式硬件的推理优化以及数据集流水线。这是一个完全远程的职位。
你将负责：
- 在实时传感器信号上开发实时检测、分类和多目标跟踪
- 将多个传感器的输入融合成一个跟踪画面
- 设计、训练并评估模型，以满足嵌入式部署的延迟和准确率目标
- 优化嵌入式硬件上的推理，包括剪枝、量化和 TensorRT
- 开发和维护数据集流水线，包括采集、标注、整理、增强、生成合成数据以及从现场数据重新训练
- 建立评估框架和指标，在实际运行条件下进行测量
- 将跟踪状态（位置、速度、身份、不确定性）传递给控制软件和操作员界面
- 按需开发后端服务和 API
- 与硬件工程师合作进行摄像头、光学和传感器选型
- 支持集成和测试，包括在目标硬件上诊断和解决故障
所需资格：
- 具有开发实时计算机视觉和多目标跟踪流水线的专业经验
- 具有在延迟、小目标和对抗条件下将检测或跟踪系统交付生产环境的经验
- 具有传感器融合经验
- 具有后端、集成和系统软件经验，除感知外
- 精通 Python 和 C++
- 具有与硬件和实时数据交互的经验
- 必须是美国公民
优先考虑：
- 具有 PyTorch、TensorRT、ONNX、DeepStream 或 GStreamer 经验
- 具有热成像、红外或多光谱成像经验
- 具有跨光电/红外、雷达或声学传感器的多模态传感器融合经验
- 具有合成数据生成和仿真到真实环境的经验
- 具有在 NVIDIA Jetson 上部署的经验
- 具有嵌入式 Linux 和设备 I/O（摄像头、云台、传感器）经验
- 具有 gRPC 或 WebSockets 经验
- 具有国防或其他安全关键型计算机视觉经验
- 具有个人机器人或硬件项目经验
查看英文原文
ABOUT THE JOB
Valkyrie Aero is seeking a software engineer to develop computer vision and sensor fusion software for an airborne mission system. Software Engineers in this role develop real-time detection, classification, and tracking software, fuse multiple sensor inputs into a single track picture, and deliver track state to the control software and operator interface. This includes model development and evaluation, inference optimization for embedded hardware, and the dataset pipeline. This is a fully remote position.
WHAT YOU’LL DO
- Develop real-time detection, classification, and multi-object tracking on live sensor feeds
- Fuse inputs from multiple sensors into a single track picture
- Design, train, and evaluate models against latency and accuracy targets for embedded deployment
- Optimize inference on embedded hardware, including pruning, quantization, and TensorRT
- Develop and maintain the dataset pipeline, including collection, labeling, curation, augmentation, synthetic data, and retraining from field data
- Establish the evaluation framework and metrics, measured under operational conditions
- Deliver track state (position, velocity, identity, uncertainty) to control software and the operator interface
- Develop backend services and APIs as required
- Work with hardware engineers on camera, optics, and sensor selection
- Support integration and test, including diagnosing and resolving issues on target hardware
REQUIRED QUALIFICATIONS
- Professional experience developing real-time computer vision and multi-object tracking pipelines
- Experience delivering detection or tracking systems to production under latency, small-object, and adversarial conditions
•Experience with sensor fusion
- Experience with backend, integration, and systems software in addition to perception
•Proficiency in Python and C++
- Experience working with hardware and real-time data
•Must be a U.S. citizen
PREFERRED QUALIFICATIONS
- Experience with PyTorch, TensorRT, ONNX, DeepStream, or GStreamer
- Experience with thermal, infrared, or multispectral imaging
- Experience with multi-modal sensor fusion across EO/IR, radar, or acoustic sensors
- Experience with synthetic data generation and sim-to-real
•Experience deploying on NVIDIA Jetson
- Experience with embedded Linux and device I/O (cameras, gimbals, sensors)
•Experience with gRPC or WebSockets
- Experience in defense or other safety-critical computer vision
•Personal robotics or hardware projects
Originally posted on Himalayas