AI Engineer
Today Secure AI removes the identity attached to your account before a request goes out. What it does not yet remove is identity inside the message — a name you typed, an address in a document, a face in a photo. Building that is this role.
It is a detection problem with an unusual constraint: the output still has to be a coherent prompt. Deleting a name gives you an answer full of blanks, so the approach is substitution and restoration, which means the pipeline matters as much as the model.
There is a second half to the job: we route every question to the model that suits it, across a blend of providers, and getting that routing right is a large part of both quality and cost.
What you would do
- Build detection and redaction for personal data across text, files, and images
- Design the tokenize-and-restore pipeline that keeps a redacted prompt answerable
- Work on model routing — matching a question to the right model at the right cost
- Build evaluation that measures what we miss, not just what we catch
What we are looking for
- Practical NLP or ML experience, particularly named-entity recognition or PII detection
- Enough engineering to ship a model into a real request path, not just a notebook
- Honesty about accuracy — our public claims depend on knowing our own error rate
- Interest in the privacy problem itself, not just the modelling
How to apply
Email us with something you have built. A repo, an app, a write-up of a problem you solved — anything that shows how you think is worth more here than a CV.
Apply for AI Engineer