Taking over a recruitment platform with 5,000+ candidates

Pachamama is a recruitment collective that places candidates in permanent jobs and freelance missions. Its whole business runs on one Bubble app. Recruiters manage assignments and candidates there, client companies follow their candidates, and candidates apply through a public job board.

In November 2024, I took over from the previous developer. I worked on the app until February 2026.

I had to take over the app without breaking the tool the whole business depends on, then deliver what the recruiters asked for.

Fast pages. The candidate page was slow, and candidate search sometimes did not load at all. I moved most of the candidate fields to a separate record, then rebuilt the search so it returns only what the results display.

Financial tracking of placements. A second recruiter per deal, referral and co-option fees, and a table that shows who earns what. I then filled in this data on the 70 existing placements.

Follow-up after a placement. The app creates the to-dos for each contract type, and sends a reminder email on the due date.

Faster rejections. With the product designer: new candidate cards, quick actions, rejection email templates, and rejecting several candidates at once.

Roles and access. Two recruiter roles with different access, a partner role for each business area, and a recruiter who leaves keeps their history in the stats.

The rest. Stats on the dashboard, Slack alerts for new assignments and placements, and data exports.

Invitations did not reach new clients. They left from the inviter's personal address, so probably went to spam. They now leave from the company domain.

Some users got a browser warning, because the email service turned secure links into insecure ones. The cause was a tracking setting.

A double click created two applications. The button now locks after the first click.

Most tickets went through a test version first, with a link the client could try. On big topics, I asked my questions before coding. For the commissions, I sent six, so the client split the work in two. When I found a data problem along the way, I explained the options and opened a ticket so it would not be forgotten.

57 tickets delivered in 15 months.

A candidate search rebuilt and live in March 2025. The database held more than 5,000 candidates in December 2025.

6 business areas merged into 3, across 7 tables, without losing any data.