Senior Computer Vision Researcher

Edgify is a global leader in edge AI solutions for retail — including item recognition and loss-prevention systems deployed in stores across the US, Israel, and Europe.

About the role

You will be a core researcher on the AI Algorithms team — the team that designs, trains, and ships the models and algorithm pipelines behind our loss-prevention and product-recognition products: object detection, multi-object tracking, item recognition, and creative use of foundation models where they move the needle.

We work AI-forward: coding agents and AI tools are part of how we build, and we keep pushing how to use them better — to improve both our products and our processes.

We are a small, senior, fast-moving team. You'll own research areas end to end and have direct influence on what we build next.

What you'll do

• Own algorithm areas in our loss-prevention stack (non-scan detection, left-in-cart, and what comes next) end to end: problem framing, dataset construction, model training, full-pipeline tuning, deployment, and monitoring in production.

• Improve real-time pipelines combining object detection, multi-object tracking, and appearance embeddings, running in live production.

• Build datasets from production data: mine hard cases and failure modes, curate tuning and test sets, drive annotation processes.

• Evaluate against business outcomes — alert rates, precision as experienced by the customer, prevented loss — not just offline benchmarks.

• Explore the frontier where it serves the product: vision-language and foundation models (auto-labeling, data mining, and beyond), action recognition, and whatever the next capability wave makes possible.

• Use and push AI tooling — coding agents and LLM-assisted research workflows — and help the team keep raising the bar on how we work.

• Adapt and deliver models for new customers and new store environments.

Requirements

• MSc in Computer Science, Electrical Engineering, Mathematics, or Physics from a leading research university, with 3+ years of hands-on deep learning experience in industry — or BSc from a leading research university with 5+ years. Exceptional candidates with less industry experience and a strong hands-on research background are encouraged to apply.

• A track record of training computer vision models and shipping them to production — models that real users or customers depended on.

• Strong Python and PyTorch; comfortable owning and evolving a substantial research codebase, not just notebooks.

• Deep understanding of modern computer vision: detection, tracking, representation learning, and the evaluation methodology to know when a result is real.

• Ownership mentality: you take a problem area and drive it — coming back with proposals and results, not a list of questions.

• Pragmatic and fast: you'd rather ship a measured improvement this week than a perfect one next quarter.

• AI-native ways of working: you already use coding agents and AI-assisted workflows day to day — or you're genuinely eager to make them central to how you work. This is how we build here.

Advantages

• Multi-object tracking or real-world video understanding experience.

• Edge inference experience: ONNX, quantization, latency/accuracy trade-offs on constrained hardware.

• Experience with vision-language or multimodal models.

• Experiment management and ML infrastructure experience (ClearML, MLflow, or similar).