Senior Computer Vision / Applied AI Engineer

Simbe Robotics•San Francisco Bay Area, CA

About The Position

Simbe is building the AI powered operating system for physical retail. Our autonomous robots and multimodal computer vision platform turn complex, constantly changing stores into accurate, actionable intelligence for leading retailers around the world. Simbe combines robotics, computer vision, machine learning, data infrastructure, and customer focused product design to help retailers improve shelf availability, price and promo execution, inventory accuracy, and store team productivity. Simbe is looking for a Senior Computer Vision / Applied AI Engineer to build production AI systems that turn store imagery into trusted retail intelligence. This role will work across dense product detection, price tag and promo tag detection, OCR, barcode decoding, product association, segmentation, visual search, model evaluation, and customer specific model improvements. The right person isa strong ML engineer, an exceptional software engineer, and a practical builder who enjoys messy real world data, rapid iteration, and measurable customer impact. Why This Role Is High Impact You will work on dense, cluttered, real world visual scenes where accuracy directly drives customer value. You will help improve core Simbe use cases such as out of stock detection, price accuracy, promo compliance, top stock, and product location intelligence. You will work on the full model lifecycle: data, annotation quality, training, evaluation, release, monitoring, and production debugging.

Requirements

5+ years of experience in computer vision, applied machine learning, robotics perception, or related production AI systems.
Strong Python experience and hands on experience with PyTorch or TensorFlow.
Experience training and evaluating object detection, segmentation, OCR, visual search, or image recognition models.
Strong understanding of dataset curation, annotation quality, model evaluation, error analysis, and experimentation.
Experience building maintainable production code and data pipelines in a Linux based environment.
Ability to balance research quality with production constraints such as latency, compute, memory, robustness, and ease of deployment.
Strong communication skills and a customer value mindset.

Nice To Haves

Experience with retail, product recognition, shelf intelligence, OCR, barcode decoding, fine grained recognition, or visual product search.
Experience with ONNX, TensorRT, CUDA, quantization, model profiling, or edge deployment.
Experience with C++, ROS/ROS2, RGBD cameras, 3D geometry, homography, calibration, or robotic perception.
Experience with FiftyOne, CVAT, Labelbox, Roboflow, or other data centric ML tools.
Experience with synthetic data generation, simulation, active learning, or automated annotation workflows.
Publications, patents, open source contributions, or production systems in relevant CV/AI domains.

Responsibilities

Build production CV models. Design, train, validate, and deploy models for object detection, segmentation, OCR, barcode localization, product recognition, shelf understanding, and other customer facing computer vision tasks.
Own dataset quality. Curate, clean, version, and analyze large real world training datasets, including hard negative mining, annotation QA, data audits, and active learning workflows.
Improve model performance. Research and implement model architecture, loss function, data augmentation, synthetic data, and evaluation improvements that increase precision, recall, latency, and robustness across customers.
Support model releases. Contribute to model validation, release gates, inference wrappers, ONNX/TensorRT exports, and production monitoring so models are reliable in customer environments.
Develop tooling. Build internal tools for model evaluation, annotation review, error mining, data visualization, dataset exports, and deployment readiness.
Partner cross functionally. Work closely with Product, Customer Success, Data, Robotics Software, and Field Operations to turn customer issues into model and pipeline improvements.
Stay current. Track modern work in open vocabulary detection, promptable segmentation, multimodal product understanding, visual search, OCR, and edge inference, and evaluate where it can create value for Simbe.