Product ML Engineer

present of AI; Paris; On-site; EUR 70,000 - 95,000; Full time

Job summary

Photoroom is seeking an ML engineer to take a focused problem from idea to a live feature within weeks, choosing between fine-tuning a model or using an external one, and writing production-ready code inside a real product.

A successful candidate ships ML features to production, evaluates models against external APIs, and owns quality after launch using PyTorch and production Python.

Qualifications

• At least two years shipping ML products.
• Solid PyTorch or similar skills.
• Practical streak that favors customer impact over academic polish.

Responsibilities

• Take a focused problem from idea to live feature within weeks.
• Decide between fine-tuning and using an external model.
• Write production-grade code inside a real product.
• Measure results after launch and own quality.

Skills

PyTorch

Production Python

ML features

Model evaluation

Tools

Diffusion models

Job description

This role asks for an ML engineer who can take a focused problem from idea to a live feature in a matter of weeks. That means choosing between fine-tuning a model and using an outside one, writing code inside a real product, and checking results after launch. A strong candidate has at least two years of shipping ML products, solid PyTorch or similar skills, and a practical streak that favors what helps customers over what is academically neat.

How it fits the AI era

The work sits right in the middle of the AI shift, since strong outside models are now easy to get and the skill is knowing when to adopt one and when to build. AI speeds up experiments and prototypes, but deciding what problem matters, judging whether an image result is good enough for a seller, and owning quality after launch stay with the person.

How it builds AI experience

Working on a photo editor that turns plain product shots into catalog-ready images gives hands-on time with image and diffusion models used by a very large user base. Someone could build toward this by fine-tuning a small diffusion model on a product photo task, comparing it with a hosted image API, and writing up how they measured which one did better.

Worth showing

• Shipped ML features to production
• Evaluating models against outside APIs
• PyTorch and production Python code
• Build versus buy decisions

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Photoroom

A generative AI photo editor that turns a plain product shot into a catalog-ready image, so online sellers do not need a studio to sell.