7 hours ago
Base Salary
$200k - $260k/yr
Responsibilities
- Design and own computer vision architecture for fine-grained catalog item identification.
- Adapt foundation vision models to specialized use cases and catalogs.
- Define category-level accuracy standards and calibrated confidence scoring.
- Build model-error feedback loops to guide labeling priorities.
- Balance broader category coverage with deeper accuracy in existing categories.
- Own the path from model research to production serving, including latency, cost, and reliability.
- Design evaluation frameworks that measure model improvements beyond training loss.
- Explain model capabilities and tradeoffs to technical and non-technical stakeholders.
Requirements
- At least 5 years of applied computer vision experience, including production systems delivered at meaningful scale.
- Hands-on experience with fine-grained or instance-level recognition of visually similar items.
- Production experience fine-tuning and deploying vision transformers or CNNs with PyTorch or TensorFlow.
- Experience adapting foundation vision models to specific use cases or catalogs.
- Experience with calibrated confidence scoring and model-serving decisions involving latency, cost, and reliability.
- Strong written and verbal communication skills and comfort providing technical leadership on an open-ended problem.
Benefits
- Hybrid role based in New York, New York.
Tech Stack
PyTorchTensorFlow