An AI hiring app that explains matches.
Chapta · full product build, live on iOS and Android
Candidates build a profile by talking to the app, in a conversation that draws out what a CV leaves off. Every match comes back with its reasoning written against it, and the rules that decide who is eligible for a role are settled in code before the model writes a word.
Overview
Chapta came to OpenKit to build a hiring platform for people tired of firing CVs into silence, and for the employers on the other side working through a pile that says very little about anyone in it.
Candidates get an app that replaces the application form with a conversation and shows them why they matched a role. Employers get a portal that creates jobs, returns shortlists with profiles redacted until both sides have opted in, and makes closing a role require a message to the people in it. Underneath both sits a matching engine we deliberately kept part rules and part model.
- A candidate app on iOS and Android, with conversational onboarding and per-job explanations of the match.
- An employer portal with shortlists, redacted profiles, reference requests and a demo environment running the real schema.
- A matching engine where the hard rules are code and the model is left to do the writing.
75%
lower cost to feed the matching engine after job ingestion moved from scraping to a jobs API.
2
app stores carrying the live candidate app, iOS and Android, iterated against real feedback.
8–12
skills drawn out of each candidate by the onboarding conversation.
3
referees a candidate can have contacted directly, with the replies surfaced against the match.
Two sides
The candidate app and the employer portal
What the candidate gets
Onboarding runs as a conversation, drawing out job history, education, skills, salary expectations, flexibility, sectors, working environment, availability and visa status.
Matches arrive with a written explanation of why this candidate fits this role, written fresh for each job.
A candidate registers interest in a role and can see where each interest stands at any point.
What the employer gets
Jobs are created with the extra fields the matching engine needs, then shortlists come back with profiles redacted until both sides have opted in.
Interest arrives with the same reasoning the candidate saw, so a first conversation starts informed.
Closing a role requires a message to the candidates in it, so nobody is left without an answer.
The app
Inside the candidate app
Approach
How the matching engine decides
Seniority filtered in code
Deciding whether someone is at the right level for a role is a hard rule with money and disappointment attached, so it never went near a model. The matching engine hard-filters seniority tiers and treats salary and location as soft criteria that shift ranking. The model writes the explanation.
Job ingestion runs on an API
Job data used to be scraped, which is fragile, legally uncomfortable and expensive at volume. Moving to a jobs API took the cost of feeding the matching engine to a quarter of what it had been, and took a whole class of breakage off the board at the same time.
A demo on the production schema
The employer demo runs against a full copy of the production schema, so a demo exercises the same matching path a customer would get. It also gave the team somewhere safe to work a fix through before it went into production.
Deletion and retention rules
A candidate profile is personal data about someone looking for work, which is about as sensitive as an application gets. OpenKit wrote and tested the deletion protocol and set a retention period before any customer asked for one.
The stack
The build and the controls
- Conversational AI onboarding
- Explainable matching model
- Mutual-interest workflow
- Real-time status tracking
- Native mobile app
OpenKit certifications
- ISO 27001
- ISO 9001, UKAS-accredited
- Cyber Essentials
Controls on this project
- UK GDPR
- UK data residency
Other engagements
What could your business do with AI?
Find out with experts who become part of your team. We uncover opportunities, get ideas working, and help your people build on the results. We reply within one working day.
Get in touch AI Audit