Decide each new role once
Today lists the roles posted in the last three days, longer over a weekend. Save one or set it aside, and it leaves the list.
NewPersonalized Fit
Checks employers’ sites automatically twice a day.
Invite only for now. Best on a laptop.

No LinkedIn Premium, no Notion, no Google Sheet. One place, free in early access.
NewPersonalized Fit
Personalized Fit learns from the roles you save, apply to and mark Not for me, and puts the roles your search finds in your order. Your model is yours alone: no AI service sees your choices, nothing reaches employers, and you can turn it off or reset it any time.
Fit 90, compared with your search. Your choices raise Fit for corporate finance roles.
Saved, Applied and the steps after it count as roles you wanted, Applied twice as much as Saved. Not for me counts as not. A role you never touched teaches nothing.
Up to 96 a role: its field, level, city, language and where it was posted, then the words of its title and posting, each hashed into one of 4,096 buckets.
Sparse logistic regression, trained with FTRL-Proximal and L1 and L2 regularisation. It is rebuilt from your current choices after each one, so an undo unlearns exactly; L1 keeps most weights at zero.
Nothing moves before 10 choices with 3 of each kind. Then your model’s weight is at most 0.25, and from 40 choices with 8 of each it can reach 0.7. The usual order keeps the rest, and Newest first is one click away.
Today lists the roles posted in the last three days, longer over a weekend. Save one or set it aside, and it leaves the list.
Pay, hours, start and language are read from the posting itself. Choose a fact to see where the posting says it. What a posting leaves out is said, never guessed.
The usual steps come ready: an online test, two interviews, the offer. Add your own, put them in order, give each a date. Choose a step for the Grachtwerk Bank internship to see it move.
Business Analyst Graduate ProgrammeHavenlicht Logistics, RotterdamFirst interviewWed 28 Oct, 09:30
Audit InternPolderveld Accountants, AmsterdamOnline testFri 23 Oct
Actuarial InternKanaalzicht Insurance, AmsterdamAppliedFri 9 Oct
NoteSpoke to a trader at the campus event.
More about your data in the privacy notice.
Straight from employers’ own career sites: about 150 employers, checked automatically twice a day. Two job boards, Magnet.me and Adzuna, fill the gaps, and their roles say so. RoleRadr follows each career site’s robots.txt and reads only what it allows.
Anyone looking for a role in Amsterdam, Rotterdam, The Hague, Utrecht or Eindhoven, in finance, software, data, product or consulting, from internships and graduate programmes to experienced roles. You choose your cities, fields and levels, and your list keeps only what matches.
RoleRadr is invite only for now. Request an invite with your address; if a place opens up, the invite comes to that address. Once invited, you sign in with a link, no password.
A small model of your own. It learns from the roles you save, apply to and mark Not for me, and puts the roles your search already finds in your order. It moves nothing until you have made 10 choices, 3 of each kind, and Newest first is always one click away.
Only you. Your model is worked out inside RoleRadr for your account alone: no AI service sees your choices, and nothing reaches employers. You can turn it off or reset it any time, and download or delete all your data in Settings.
Yes. Every part of the app is there, and each role opens as its own page. It’s best on a laptop, where a posting’s facts sit beside its words.
No. You apply on the employer’s own page. Mark a role applied and the Tracker keeps it with the date of each step, through to an offer.
A role written in Dutch, or asking for it, carries an orange ring and the words In Dutch. Your search can leave out roles that require Dutch, and the language filter can hide every role in Dutch.
Yes. RoleRadr reads only public postings, sends people to your own page to apply, and stops reading a site whose robots.txt asks it to. How to have a posting or your site removed.