> Markdown version of [/jobs/ext/4826258-senior-computer-vision-engineer](https://www.wearedevelopers.com/jobs/ext/4826258-senior-computer-vision-engineer). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Senior Computer Vision Engineer - **Company:** Arrive AI Inc. - **Location:** Fishers, IN, United States - **Experience:** Expert - **Contract:** Permanent contract - **Skills:** Application Programming Interfaces (APIs), Amazon Web Services, Amazon S3, Computer Vision, Microsoft Azure, Unix, C++ (Programming Language), Computer Programming, Data Validation, Data Files, DevOps, Protocol Buffers, Inter-Process Communication, Python (Programming Language), Machine Learning, Language Modeling, Raspberry Pi, Cloud Services, Software Deployment, Data Streaming, Management of Software Versions, WebRTC, Edge Inference, Pytorch, Test Scripts, RTSP, Containerization, Kubernetes, Information Technology, Low Latency, ONNX (Open Neural Network Exchange) Format, Triton Inference Server, Machine Learning Operations, TensorRT, Drift Detection, Api Design, NVIDIA Jetson, Grpc, Docker - **Published:** November 21, 2026 - **Apply:** https://www.indeed.com/viewjob?jk=a28acf454b0d267c ## About the Role * Education and experience: A master's degree in Computer Science, Computer Vision, Machine Learning, Electrical Engineering, or a related field with 3 to 5 years of industry experience, or a PhD with 3 or more years of industry experience. We weigh hands-on work with real cameras and edge hardware heavily. * End-to-end ownership: A track record of shipping a vision system from camera to production deployment, not only training models. * Model development: Classical and modern computer vision, with experience evaluating, selecting, training, and fine-tuning classifiers, CNNs, vision transformers, object detectors, and vision-language models in PyTorch. * Edge inference: TensorRT and ONNX, including engine builds, quantization, and accuracy checks after conversion. * Media pipelines: NVIDIA DeepStream and GStreamer. * Serving and APIs: Triton Inference Server, gRPC, and modern API-driven development workflows. * Edge hardware: NVIDIA Jetson (Orin, Thor) and other edge compute such as Raspberry Pi. * Programming: Strong C++ (C++17 or later) and Python. * Data on real hardware: Confidence collecting data on edge devices, including sampling, curation, labeling, edge-case analysis, and adjusting camera and lighting conditions. * Containers and DevOps: Docker, Kubernetes or K3s, CI/CD pipelines, and working familiarity with AWS or Azure and S3-compatible object storage. * Communication: Clear writing and presenting for technical and non-technical audiences, and a habit of documenting as you build. * Learning and flexibility: Quick to pick up new tools, platforms, and problems in a fast-moving startup. ## Description We're seeking a Senior Computer Vision Engineer to own the Arrive Point vision pipeline end-to-end: from the camera, through inference on the edge device, to the events that ArriveOS acts on and the data that reaches the cloud. You'll decide which models we run, train and optimize them, deploy them to NVIDIA Jetson hardware, and keep them accurate in the field. This is a hands-on senior role with a bias toward the edge device. You'll work directly in C++ and Python with NVIDIA's DeepStream, GStreamer, Triton Inference Server, TensorRT, and related frameworks. You'll be familiar with the cloud side and work closely with our platform team, who manages our operating system and cloud services. What You'll Do * Own the vision pipeline: Own the Arrive Point vision stack from camera ingress through inference to the events ArriveOS consumes and the data that reaches the cloud. * Build edge pipelines: Develop and maintain C++ pipelines with NVIDIA's DeepStream and GStreamer on Jetson Orin and Thor, using Triton Inference Server and TensorRT. * Select and train models: Evaluate, select, train, and fine-tune the right model for each task, from binary classifiers and CNNs to vision transformers, object detectors, and vision-language models. * Optimize for the edge: Export models through ONNX to TensorRT, build per-platform engines (FP16, INT8), and verify accuracy, latency, and memory on the device. * Collect data on real hardware: Sample, curate, and label data from deployed cameras, fine-tune their configurations, and address edge cases. * Run the model lifecycle: Maintain the training and release pipeline, including dataset versioning, CI-gated evaluation, ONNX and TensorRT parity checks, robustness testing, drift monitoring, and retraining. * Integrate with ArriveOS: Work with the platform team on the gRPC contracts between the vision service and ArriveOS, our Rust-based edge orchestrator, including event debouncing, health reporting, and inter-process communication. * Stream video: Maintain live video egress over WebRTC for remote viewing and operations. * Deploy and operate: Containerize and deploy vision services to edge devices with Docker and K3s, including cross-builds, hardware-in-the-loop tests, and simulated test environments. * Work with the cloud: Collaborate with our cloud and platform team members on object storage, model publishing, and data flows between the edge and the cloud. * Protect camera data: Design with privacy and security in mind for video that can contain personal information (HIPAA familiarity and "on-prem only" experience are a bonus). * Document and communicate: Write architecture decision records, technical guides, and test reports, and present results and trade-offs to engineers, leadership, and partners., * Specific models: YOLO, TinyViT, NVIDIA PeopleNet, and the NVIDIA TAO Toolkit. * WebRTC: Live video with WebRTC, including SFUs, WHIP, and TURN and STUN. * Inter-process communication: Unix domain sockets, shared memory, and message framing with FlatBuffers or Protocol Buffers. * Camera hardware: Image sensors, optics, housings, GigE, MIPI CSI-2, USB and V4L2, Intel RealSense, and time-of-flight sensors. * Other languages: Familiarity with Rust-based architectures, and Go. * Additional perception: OCR and depth sensing. * MLOps: Dataset versioning, drift monitoring, and label quality analysis. * Video integration: RTSP, ONVIF, and video management systems. * Domain: A background in logistics, delivery, retail, or security video. Additional * Expect to travel up to 5 days per month. * Responsibilities include all other duties as assigned. * At this time, we are not able to provide visa sponsorship for this position.