Skip to content
AwakastLet's talk

Confidential HRTech client, recruitment platform

Recruiters now review 4.1× more candidates a day

We rebuilt a recruitment product around AI screening on a privately hosted model, with interview questions for each candidate and billing built in. It went from concept to a market-ready first version in three months.

From an idea

faster candidate screening
4.1×
profiles reviewed per recruiter per day
12 to 49
from concept to market-ready MVP
3 months

Where the client started

The client runs a recruitment product. Recruiters were held up by reading CVs by hand and preparing the same follow-up questions again and again. To stay competitive, the product needed a full rebuild with AI at its core. The new version had to take in candidates in several languages, screen far more of them and give recruiters clear oversight of every AI-assisted decision. Candidate data had to be processed privately.

What we built

We rebuilt the product end to end. A privately hosted model scores each CV using prompts written for the specific role. A Smart Questions module prepares targeted interview questions for each candidate. When the model is not confident enough, the profile goes to a recruiter for mandatory review. Intake, scoring and the recruiter screens run as separate services on Kubernetes, so the platform grows with the load. We also added subscriptions and billing with Stripe, plus a dashboard and monitoring for daily operations.

Deliverables

  • Private AI scoring with prompts tailored to each role

  • Smart Questions module for targeted interviews

  • Subscriptions and billing

  • Operations dashboard and monitoring

How it works

  1. Candidates come in, in several languages.

    A separate intake service takes in CVs and passes them on for scoring.

  2. A private model scores each profile.

    Scoring uses prompts written for the role, and candidate data is processed on private infrastructure.

  3. Low confidence means a person reviews.

    Scores below the threshold go to a recruiter for mandatory review.

  4. Recruiters get questions, not only scores.

    Smart Questions suggests what to ask each candidate in the interview.

Results

faster candidate screening
4.1×
from concept to market-ready MVP
3 months
  • Average reviewed profiles12 before, 49 after

  • Recruiters moved from scanning CVs by hand to guided decision support

  • Structured scoring made candidate evaluation more consistent

  • The product launched on an architecture built for enterprise customers

The client's name, some model documentation and client-specific thresholds are not shown.

3 months to a market-ready MVP

  1. Month 1

    Scope and architecture

    Set the product boundaries and the data flow for an AI-first applicant tracking system.

  2. Month 2

    Core build

    Built scoring, ranking, Smart Questions and the recruiter panels.

  3. Month 3

    Launch readiness

    Hardened the system, added controls and supported the commercial rollout.

Technology

  • Python
  • Mistral (privately hosted)
  • NLP
  • React
  • Node.js
  • PostgreSQL
  • Kubernetes
  • Stripe

Show us the part of hiring your recruiters still do by hand.

Tell us what you're building