- Work from
- Anguilla, Antigua and Barbuda, Aruba, Bahamas, Barbados, Belize, Bermuda, Bonaire, Sint Eustatius and Saba, Canada, Cayman Islands, Costa Rica, Cuba, Curaçao, Dominica, Dominican Republic, El Salvador, Greenland, Grenada, Guadeloupe, Guatemala, Haiti
- Contract
- Full-time
- Posted
- 16 November 2026
- Apply by
- 15 January 2027
As founding ML engineer, you'll own the identification architecture, data strategy, evaluation standards, and production serving path — end to end, from foundation models fine-tuned toward a specific catalog.
This is a remote position.
On behalf of an AI commerce tech startup · Remote in North America · Full-time
The build:Our client is an early-stage commerce tech company building production computer vision for precise item identification and pricing.
The problem is fine-grained: brand, model, edition, variant. "Close" isn't good enough.
Responsibilities
- Design and own the CV architecture for fine-grained item identification — starting from foundation vision models, fine-tuning toward the company's catalog.
- Define what "accurate enough" means per category; build calibrated confidence scoring so the product says "we're not sure" instead of guessing.
- Build the feedback loop between model errors and what gets labeled next, with the labeling lead.
- Decide where to invest: broader category coverage vs. deeper accuracy on today's categories.
- Own the serving path from research to production — latency, cost, reliability at scale.
- Represent the model's capabilities and limits company-wide — including with investors and customers.
Requirements
- 5+ years applied CV, with at least one system shipped to production at meaningful scale.
- Hands-on fine-grained / instance-level classification — a problem where "close" isn't good enough.
- Fluency in PyTorch or TensorFlow; fine-tuning and deploying vision transformers or CNNs in production.
- Evaluation frameworks for vision models — you know when a model isactuallygetting better.
- Comfortable being the most senior technical voice on an open-ended problem with no playbook.
- Strong communication — you explain trade-offs to non-technical stakeholders.
Nice to have
- resale/marketplace (StockX, GOAT, Vinted, Rebag, The RealReal) or visual search (Pinterest Lens, Google Lens, Syte)
- active learning / human-in-the-loop pipelines
- low-latency API serving at scale
- first-ML-hire startup experience
F
S
P
L
K
Discussion
No comments yet. Ask a question or share a tip about this role.