Edge Computer Vision Engineer

Design, build, and deploy the visual intelligence that powers Rownd's autonomous drones and edge assets. This is hands‑on work with real cameras, sensors, compute, drones, and field constraints. If you want to see your code fly, you belong here. This role is available only in Research Triangle Park, North Carolina. Candidates must already be local or willing to relocate. This is a hybrid role with expected worksite time a few days per week. Compensation Competitive salary + equity, DOE Location Research Triangle Park (RTP), North Carolina; candidates must be local or willing to relocate Worksite Expected on‑site in RTP a few days per week for hands‑on hardware, sensors, drones, and team integration Travel 5-6 cumulative domestic travel weeks per year across team sessions, client sites, and conventions About Rownd We build full physical autonomy from the ground up. Rownd is a fast‑paced, high‑execution technology and hardware startup building 100% autonomous drones, UxS platforms, and sensors for defense and commercial sectors. We do not just write software or integrate off‑the‑shelf parts; we build the full drone from the ground up. We operate with extreme autonomy, an aggressive bias for action, and a just‑do‑it mindset. We do not write software to sit in the cloud. We write software to make physical things navigate, sense, and act in the real world. Because we build the full hardware platform, you have complete ownership over the vision pipeline: cameras, sensors, processors, models, deployment stack, and field feedback loops. This role is deliberately hands‑on; you should expect to spend real time with hardware on a workbench, at a worksite, and in field‑testing environments. Core focus Edge model design and deployment Full-stack hardware and sensor selection Research, evaluate, and select camera modules, lenses, optical sensors, and onboard edge compute processors to meet mission requirements. Implement, optimize, and deploy state‑of‑the‑art vision models, including YOLO variants, CLIP, DINOv2, and newer backbones worth adopting. Small and fast deployments Fine‑tune, compress, and quantize models for constrained hardware environments across tracking, classification, and segmentation. CUDA and non‑CUDA deployment Target NVIDIA Jetson‑class CUDA devices and non‑CUDA NPUs such as Hailo, Rockchip, STM32, or similar edge accelerators. Self‑improving systems Design feedback loops so deployed assets can flag edge cases, gather raw data, and improve over time. Robotics, systems, and operations Use simulation, generative tools, and programmatic pipelines to create high‑fidelity synthetic training data when real‑world data is scarce. Write clean, modular, performant C++ and Python that integrates model outputs into drone navigation and control pipelines. Use AI coding assistants such as GitHub Copilot, Cursor, and LLMs to draft, document, debug, and iterate rapidly. Attend daily team standup at 10:30 AM EDT with camera on and ready to actively engage. US citizen eligible to obtain and maintain a U.S. Security Clearance due to defense work and FCL requirements. Hands‑on experience deploying computer vision models to physical hardware and optimizing for INT8/FP16, memory, thermal, and power limits. High proficiency in PyTorch or TensorFlow, OpenCV, TensorRT, CUDA, C++, and Python. A proven just‑do‑it mentality: prototype, test, break it, fix it, and keep moving in ambiguous environments. Located in Research Triangle Park, North Carolina, or genuinely willing to relocate for a hybrid worksite role. Willingness to travel 5-6 cumulative weeks per year for on‑sites, hardware integration, and client field testing. Culture, trust, and expectations High trust, exceptionally high expectations. Figure it out You will not be handed a step‑by‑step playbook. You are expected to think outside the box, try wild new things, conduct rapid research, and do what it takes to accomplish the mission. Startups are fast and chaotic. Ambiguity, shifting requirements, and rapid change should push you to adapt and build, not freeze. Positive and collaborative We are a tight‑knit team. We look for people who are additive, collaborative, positive, and genuinely great to work with. No brilliant jerks. Hard work and real rest We work intensely when we are on, and we also expect you to actually take time off, unplug, and recharge. Defense mission alignment You must be willing and happy to provide software, models, and hardware to the Department of War. If you have moral reservations around supplying the US military and government, this role is not for you. Jack of all trades, master of SOME What are you absolutely great at? Examples: custom CUDA kernels, synthetic data pipelines, physical NPU integration, sensor selection, or edge model compression. Where do you need or want to learn? Tell us your gaps, what you are excited to master next, and what newer stacks you are dying to play with. What's in it for you Real hardware, real autonomy, real ownership. You will build defense‑grade, audit‑ready AI systems at a rapidly expanding startup. Rownd will equip you with the hardware, compute, and AI developer tools needed to move fast. Competitive salary plus meaningful equity gives you direct ownership in Rownd's growth. Full-time Hybrid in RTP, NC Daily 10:30 AM EDT standup U.S. citizenship required #J-18808-Ljbffr Rownd