Founding CV ML Engineer — Real-World Visual ID System

Moe; New York (NY); On-site; USD 200,000 - 260,000; Full time

Job summary

Moe is hiring for a senior computer vision lead who will define and own the architecture for fine-grained item identification across our catalog. You will start from foundation vision models, fine-tune toward Moe's products, and ship a production-ready system that can accurately distinguish brand, model, edition, and variant.

You will set the accuracy bar, design calibrated confidence scoring, and build a scalable serving path.

Qualifications

• 5+ years in applied computer vision, with at least one system shipped to production at meaningful scale.
• Hands-on experience with fine-grained/instance-level classification, not just general object detection.
• Fluency in PyTorch or TensorFlow and experience deploying vision models in production.
• Experience designing and running evaluation frameworks to measure real model improvement, not just lower loss.
• Strong written and verbal communication to explain tradeoffs to stakeholders, including investors.

Responsibilities

• Design and own the computer vision architecture for fine-grained item identification — brand, model, edition, variant.
• Define what 'accurate enough' means per category and build calibrated confidence scoring.
• Build feedback loops between model errors and labeling priorities with the labeling lead.
• Decide where to invest: broader category coverage vs. deeper accuracy on current categories.
• Own the model serving path from research to production — latency, cost, and reliability at scale.
• Represent the identification model's capabilities to the company and investors when needed.

Skills

Computer vision

Fine-grained classification

Production deployment

PyTorch

TensorFlow

Model evaluation

Strong communication

Tools

Transformers

Job description

Moe is hiring for a senior computer vision lead who will define and own the architecture for fine-grained item identification across our catalog. You will start from foundation vision models, fine-tune toward Moe's products, and ship a production-ready system that can accurately distinguish brand, model, edition, and variant.

You will set the accuracy bar, design calibrated confidence scoring, and build a scalable serving path.