EyePop.ai is an SDK-first computer vision platform. Using our Python or Node/TypeScript SDK, developers and AI coding agents compose production-grade detection, tracking, pose, and OCR pipelines and deploy them in under an hour, without a dedicated ML team. EyePop also powers Physical AI Vision: on-device computer vision for robots, drones, and autonomous systems, running the same pipeline definition across NVIDIA Jetson Orin, Qualcomm Snapdragon Dragonwing NPU devices, GPU servers, or CPU-only devices, in the cloud, on-premise, or at the edge.

EyePop.ai was founded in 2023 by Brad Chisum (CEO, previously co-founded Lumedyne Technologies, acquired by Google in 2014), Torsten Schulz (CTO, 20+ years scaling startups including Jelli, Inc. and Yahoo!), and Andy Ballester (CPO, co-founder of GoFundMe). The company's mission is to make AI-powered vision models accessible, seamless, and completely within your control, so you can build, deploy, and own your vision data solutions with confidence.

No dedicated ML team is required, but EyePop.ai is built SDK-first for software developers and AI coding agents, not as a point-and-click tool. The primary interface is the Python and Node/TypeScript SDK; a visual Workflow Designer is also available, and a Claude Skill lets AI coding agents build and configure inference pipelines directly inside Claude Code.

Yes. EyePop.ai offers a library of pre-built models and Abilities for common use cases like object counting, product detection, and measurement, including dense image captioning, zero-shot detection, and vision-language models (VLMs). These can be used as-is, fine-tuned with your own data, or combined with custom-trained models through Pops, EyePop's composable inference pipelines. Browse examples in the Abilities Hub.

Yes. Your data is your own; EyePop.ai never repurposes, shares, or analyzes it outside your specified use case, and EyePop.ai is HIPAA certified. On IP: EyePop.ai does not yet hold a company patent. An application is in preparation covering the platform's model composability approach, the Pops pipeline architecture.

EyePop.ai's Self-Service Training lets you train a custom AI vision model in a few hours, automated end-to-end through the SDK or driven directly from the dashboard.

Most users can end-to-end train a custom AI-powered vision model in just a few hours. The platform automates model optimization, training, and deployment so you can move quickly from testing to production.

Our Self-Service Training platform includes auto-labeling, which uses AI to automatically generate high-confidence annotations on your dataset, accelerating the labeling process and reducing manual effort.

To train a custom AI‑powered vision model on EyePop.ai, you’ll need a set of images or video frames that represent the objects or scenarios you want the model to detect. These images should be clear, high‑quality, and relevant to your use case.

Many users successfully train a model with as few as 200 labeled images. More images generally lead to higher accuracy, but you can always start small and add more data later to improve performance.

You can upload JPEG, PNG, or MP4 video files. For video, the platform will extract frames automatically for labeling and training.

No. EyePop.ai provides a built‑in labeling interface so you can tag objects after upload. You can also upload pre‑labeled datasets through our SDK if you already have annotations.

Initial accuracy depends on the quality and quantity of your dataset. Many users achieve strong results with as few as 200 images, but you can always add more data and retrain to improve performance over time.

Yes. Models can be retrained or fine‑tuned at any time. As you collect more data, simply upload new images, label them, and start a new training run.

Our Self-Service Training platform supports an iterative approach—train, deploy, test—powered by a continuous learning system that helps you identify edge cases and retrain to make your model smarter with every cycle.

You retain full ownership and control of your models, datasets, and outputs as long as your account is active. Unlike some platforms, we don’t claim rights to your trained models.

EyePop.ai offers three flexible deployment paths tailored to your workflow and scale:

• Cloud — Production ($200/mo): Ideal for development, prototyping, and low-volume apps. Includes 4,000 compute units per month on general-use infrastructure, with additional overages billed transparently at $0.05 per unit.

• Cloud — Enterprise (starts at $800/mo): Built for organizations running centralized inference at scale. Features dedicated, always-up, low-latency servers with no overage fees, custom SLAs, and multi-seat collaboration.

• On-Premise AI Application Runtime (volume-tiered): designed for local deployments, from CPU and NVIDIA GPU servers to edge hardware including NVIDIA Jetson Orin (Thor, NX, and Nano) and Qualcomm Snapdragon Dragonwing NPU devices, including Dragonwing IQ9 and Snapdragon X Elite for Windows. Pricing scales down on a per-device basis as your volume grows. Video data is processed entirely on your local hardware, keeping data local to comply with strict security requirements while transmitting only lightweight metadata.

For a full breakdown of feature matrices, volume discount tiers, and to choose a plan, visit our Pricing page. You can also contact our team for a custom architecture review.

Our API is REST‑based, meaning it works with virtually any language or framework, including Python, Node.js, Java, C#, and more.

We offer dedicated SDKs for Python, Node, and React, along with code examples to accelerate your development. Explore developer resources.

Yes. Trained models can be deployed locally with our On-Premise AI Application Runtime, running on CPU, NVIDIA GPU servers, or edge hardware like NVIDIA Jetson Orin and Qualcomm Snapdragon Dragonwing NPU devices, the same Pop definition runs unmodified across targets. This eliminates the need to run your data through a cloud service, ensuring more privacy, lower latency, and near-zero bandwidth.

Yes. Once trained, your models can be accessed via API or SDKs, or integrated into your app, dashboard, or workflow with our easy to use JSON response. See developer documentation.

No. EyePop.ai handles all compute, scaling, and infrastructure for you. The Production Plan offers faster inference, dedicated compute resources, and team support for production deployments.