Skip to content
Solutions · AI Interview Software

Self-Hosted AI Interview Software for Screening Calls

AI interview software from NeuralChainAI runs first-round screening calls with candidates by phone, asks your screening questions and writes a structured summary to the ATS for the recruiter. Speech-to-text, the language model and text-to-speech all run self-hosted on open-weight models in your own cloud account or on your servers, so call audio and transcripts stay inside the firm's environment.

Book a Voice Screening Strategy Session Free 30-minute call · mutual NDA included
100%Target: call audio, transcripts and summaries stored on the firm's own infrastructure.
24/7Candidates take the screening call when it suits them, including evenings and weekends.
10Business days of notice NYC candidates must get before an automated employment decision tool is used, under DCWP rules.
Outcomes

What a Self-Hosted Screening Call Does

Six parts of a first-round screen, handled by a voice agent your recruiters configure and review.

Places and answers calls

Calls candidates at the booked time or takes inbound calls from a job ad, in a natural voice, and introduces itself as an AI assistant for your firm.

Asks your screening questions

Works through the questions for the role, such as availability, rate, location, work authorisation and must-have skills, and follows up on vague answers.

Writes a summary to the ATS

Answers, a short summary and the full transcript land on the candidate record in Bullhorn, Loxo, Recruit CRM or Greenhouse for the recruiter to review.

Books a recruiter call

At the end of the screen, the agent offers a slot from the recruiter's calendar, so every candidate speaks with a person next.

Self-hosted voice stack

Speech-to-text, an open-weight LLM and text-to-speech run on your GPUs or in your cloud account, and audio is stored under your own retention policy.

Disclosure, consent and review

Each call opens with an AI disclosure and recording consent. The agent records answers, and recruiters make the screening decision.

Why Self-Hosted Voice

Why Firms Run AI Screening Calls on a Self-Hosted Voice Stack

Paradox, Manatal's AI Interviewer add-on and the voice AI features in platforms such as Greenhouse run AI interviews on the vendor's cloud and its model providers. Firms choose a self-hosted voice stack when one of these applies:

1 Client contracts that set where candidate audio and transcripts may be processed.
2 Confidential searches, where a candidate's interest in a role is itself sensitive.
3 Call recordings kept under the firm's own retention and deletion policy.
The Self-Hosted Answer

Screening calls on a voice stack you host.

Speech-to-text, the language model and text-to-speech run on open-weight models in your environment, on the same stack as our custom voice agents. Audio, transcripts and summaries are written to your storage and your ATS, under your access controls.

Speech-to-text, LLM and text-to-speech all self-hosted
Summaries written to your ATS
A recruiter reviews every call
Inside the Deployment

How We Build Self-Hosted AI Interview Software

Six pieces, from the screening script to the audit log, all running in the firm's own environment.

1

A screening script per role family

We turn your recruiters' screening questions into scripts per role family, with knock-out questions, follow-up prompts and the ATS field each answer fills.

2

Telephony and scheduling

Calls run over your telephony provider or SIP trunk. Candidates pick a slot from a link, or the agent calls at the booked time, with retries and voicemail handling.

3

Self-hosted speech-to-text

An open-weight speech recognition model runs on your GPUs, with your role vocabulary, such as skills, certifications and client names, supplied as context so transcripts read correctly.

4

Self-hosted LLM and text-to-speech

An open-weight LLM served with vLLM runs the conversation and writes the summary, and a self-hosted text-to-speech model gives the agent its voice. We tune latency for natural turn-taking.

5

ATS write-back

Answers, the summary, the transcript and a link to the recording are written to the candidate record, and the recruiter gets a review task.

6

Disclosure, consent and audit logs

Each call opens with an AI disclosure and recording consent. Calls, transcripts and recruiter decisions are logged, so the firm can answer candidate requests and support bias audits where a tool falls under rules such as NYC Local Law 144.

Start Today

Talk to a Voice Screening Expert

Bring one role family you screen often and the questions your recruiters ask. We will map the script, the hosting for the voice stack and the ATS write-back, and give a directional cost for your firm.

Book a Strategy Session →
Ask us about
Self-hosted speech-to-text and text-to-speech
Screening scripts per role family
Telephony, SIP and scheduling
ATS write-back of answers and transcripts
Disclosure, consent and recording retention
Bias-audit logging for screening tools
Own the Capability

When a Self-Hosted Voice Stack Fits

Hosted AI interview tools suit high-volume hiring where clients accept vendor-cloud processing of candidate audio. A self-hosted voice stack fits these cases:

Confidential searches — a candidate's interest in a role stays inside the firm.
Client data terms — contracts that set where candidate audio and transcripts may be processed.
Your retention policy — recordings stored and deleted on your schedule.
One voice stack, several uses — the same deployment can run candidate reminders and availability checks.
Your ATS fields — answers land in the structured fields your recruiters already use.
Steady economics — a one-time build plus hosting, with call cost set by your own compute.

A self-hosted voice stack gives recruiters a first-round summary for every candidate and keeps the audio, transcripts and decisions inside the firm.

Questions

Frequently Asked Questions on AI Interview Software

AI interview software runs part of a job interview with an AI system, usually a first-round screen by phone, video or chat, and records the answers for a recruiter. Ours handles phone screening calls on a self-hosted voice stack: it asks the role's screening questions, follows up on unclear answers and writes a structured summary to the ATS. A recruiter reviews every summary and makes the decision.

All three models run in the firm's environment: speech-to-text, the language model and text-to-speech, each an open-weight model on GPUs in your cloud account or on your servers. Call audio, transcripts and summaries are stored there too. Only the phone connection runs through your telephony carrier.

Yes. Each call opens with a disclosure that the caller is an AI assistant for the firm and asks for consent to record. Candidates can ask to speak with a recruiter at any point, and the agent books that call.

It depends on where candidates live and how the output is used. In New York City, a tool that substantially assists hiring decisions needs an independent bias audit within one year before use and 10 business days of candidate notice under Local Law 144. Colorado's SB 26-189 adds notice and explanation duties from January 1, 2027, and Quebec's private-sector privacy act requires telling a person when a decision is based exclusively on automated processing. Call-recording consent rules vary by state and province. Your counsel sets the policy, and the system logs what an audit needs.

Bullhorn, Loxo, Recruit CRM and Greenhouse through their APIs, plus other systems with an open API after scoping. The summary, answers, transcript and recording link land on the candidate record, and the recruiter gets a review task.

GPU capacity for the three voice models, the number of role families and scripts, call volume and the ATS integration. The build is a one-time cost, then hosting and an optional flat retainer. The models run on your compute, so call cost follows your own infrastructure.

Ready to Run Screening Calls on Your Own Voice Stack?

A 30-minute strategy session. We will scope the script, the self-hosted voice stack and the ATS write-back, with a directional read on cost.

Discuss Your Project