Skyllect

Building JobTalk AI's Voice Screening and Recruiting Workflows

Voice-first screening that moves candidates instead of queueing them

JobTalk AI screens candidates by voice. The bottleneck in hiring is rarely the decision - it is the days spent arranging and running first-round calls. JobTalk runs that round conversationally, then pushes structured results back into the systems recruiters already work in. Skyllect built the entire platform, from the backend and web app to the voice agents and model fine-tuning.

www.jobtalk.ai
The JobTalk AI homepage
Sector
Recruiting technology
Surface
Voice and web
Users
Recruiters and candidates
Focus
Screening and workflow

The first round is where hiring stalls

Early-stage screening is high volume and low variance. The same questions get asked hundreds of times, calls are hard to schedule across time zones, and strong candidates go cold while waiting for a slot.

Automating it is not simply a matter of a chatbot. A screening conversation has to feel like a conversation, has to be consistent enough to support structured human review, and has to land in the ATS as structured data rather than a transcript nobody reads.

What we built

End-to-End Build

Skyllect designed and built the whole product: backend services, the recruiter and candidate web app, the voice agents and the integrations.

Voice Agents That Handle Real Calls

Fine-tuned language models and voice agents built for how people actually talk, including interruptions, pauses and off-script answers, and tested against a wide range of real-world questions and answers.

Voice-First Screening

A spoken first-round interview candidates can take when it suits them, asking a consistent set of questions per role.

Candidate Engagement

Follow-up and status communication that keeps candidates warm through the gap between applying and hearing back.

Scheduling

Availability matching and booking for the rounds that do need a human, without the back-and-forth email thread.

ATS and HRIS Integration

Results written back as structured fields into the systems of record, so screening output is usable rather than archived.

Recruiter Analytics

Funnel, pipeline and outcome reporting across roles, so the process itself can be measured and adjusted.

Built with

  • Next.js
  • TypeScript
  • Node.js
  • Speech processing
  • Voice agents
  • LLM fine-tuning
  • PostgreSQL
  • ATS/HRIS APIs
  • Cloud infrastructure

What it changed

  • First-round screening runs without a recruiter present on the call
  • Structured screening questions support recruiter review; they do not establish fairness or remove the need for bias evaluation
  • Screening results land in the ATS as structured data, not transcripts