Autonomous AI Recruitment Pipeline

From CV to hire decision,
fully autonomous.

TalentForge runs your entire top-of-funnel hiring pipeline autonomously — CV screening, real-time AI interviews, and explainable evaluation reports. You stay in control at every critical decision point.

End-to-end autonomous pipeline

1 CV Upload
2 AI Screening
3 Schedule Negotiation
4 AI Interview
5 Evaluation Report
6 ATS Sync

Human-in-the-loop checkpoints at every critical stage. AI assists, you decide.

Built for how hiring actually works

Not another ATS. An autonomous pipeline that does the work while keeping you in control of every decision that matters.

RAG-Powered CV Screening

AI extracts skills, experience, and education from CVs, then scores candidates against your job description and company knowledge base using retrieval-augmented generation. Weighted criteria you control.

Real-time AI Interviews

30-minute structured interviews via WebRTC with live speech-to-text and text-to-speech. STAR framework probing with real-time follow-ups. Avatar lip-sync for natural conversation. English, Japanese, and Vietnamese.

Explainable Evaluation (XAI)

Every hire/pass recommendation comes with chain-of-thought reasoning, per-competency STAR analysis, and risk flags. Defensible decisions you can explain to candidates and stakeholders.

Smart Schedule Negotiation

AI proposes available time slots from Google Calendar. Candidates pick a slot or suggest alternatives. No back-and-forth email chains — the system handles negotiation autonomously.

Human-in-the-Loop at Every Stage

The pipeline pauses at configurable checkpoints: low scores, risk flags, quarantined CVs. You approve, reject, or override. AI never makes final decisions without your sign-off.

Conversational HR Assistant

Chat with your hiring data. Ask "who are the top candidates for the backend role?" or "approve this candidate" — the AI assistant executes actions directly on your pipeline.

How the AI actually works

No black box. Every decision is traceable, explainable, and auditable.

1

Screen & Score

CVs are parsed, PII is stripped before the AI sees anything. The LLM scores against your job description + knowledge base using weighted criteria you define.

technical_skills: 40% · experience: 30%
education: 15% · culture_fit: 15%
2

Interview (STAR)

30-minute structured behavioral interview. AI detects STAR gaps in real-time and probes deeper. Perception monitoring tracks engagement. Full transcript saved.

Situation → Task → Action → Result
Gap detected? → Follow-up probe
3

Evaluate & Explain

A separate AI model evaluates the interview + profile against rubrics. Produces chain-of-thought reasoning, risk flags, and a recommendation with full justification.

score: 82 · recommend: HIRE
reasoning: "Strong system design..."

Two different AI models are used: one for screening/interview, another for evaluation. This multi-model approach prevents single-model bias and enables cross-validation.

Built for teams that hire at scale

Whether you're processing 100 or 10,000 applications per campaign, TalentForge handles the top-of-funnel while you focus on final-round decisions.

  • Custom branded portal — Candidates apply at yourcompany.portal.talentforge.cloud
  • Knowledge base — Upload company docs, rubrics, and policies. AI uses them to score candidates contextually via RAG
  • Custom scoring rubrics — Define weighted criteria per role and level. AI suggests rubrics from your JD automatically
  • Role-based access (RBAC) — Recruiters, hiring managers, and admins each see what they need
  • Multilingual — AI interviews in English, Japanese, and Vietnamese with native-quality speech
  • Configurable thresholds — Set pass/fail scores per level. Borderline cases route to human review automatically
JD
Senior Backend Developer
Engineering · 2 positions
Active
142
Applied
28
Interviewed
5
Shortlisted
AI Evaluation — XAI Report
Technical Skills
88
Experience Match
92
STAR Completeness
75
HIRE Strong system design demonstrated in STAR responses

Security and compliance by design

PII never reaches the AI unprotected. Built on AWS with tenant isolation, encryption, and regulatory compliance baked into the architecture.

PII Pseudonymization

Names, emails, and personal data are replaced with pseudonyms before any AI model sees the CV. Original data encrypted with AES-256.

Tenant Isolation (RLS)

PostgreSQL row-level security ensures complete data separation. Each tenant's queries physically cannot access another tenant's data.

GDPR + EU AI Act

Right to erasure, consent management, and explainable AI decisions (XAI). Human-in-the-loop satisfies EU AI Act high-risk requirements.

Injection Defense

Prompt injection detection catches hidden text in PDFs, adversarial instructions, and XML delimiter attacks. Flagged CVs go to quarantine review.

Bias Monitoring

Continuous fairness metrics tracking across scoring outcomes. Alerts when patterns suggest bias across protected characteristics.

Full Audit Trail

Every pipeline action logged with timestamp, actor, and context. Immutable audit records for compliance reporting.

Infrastructure

AWS ECS Fargate in private subnets, WAF, encrypted RDS, Secrets Manager. Zero public-facing backend components.

Fits into your existing stack

TalentForge connects to the tools your team already uses. Full REST API for custom integrations.

📅
Google Calendar
Free/busy sync + auto-scheduling
🔑
SSO (OIDC)
Keycloak, Okta, Azure AD
💬
Slack + Email
Pipeline event notifications
🔄
ATS Webhooks
Greenhouse, Workday, custom

Full REST API with JWT authentication. Webhook-based event system for real-time pipeline updates.

Built with

Python + FastAPI LangGraph LiveKit (WebRTC) PostgreSQL + pgvector AWS ECS Fargate Next.js

Ready to automate your top-of-funnel?

See how TalentForge handles screening, interviewing, and evaluation for your open roles. We'll set up a demo with your actual job descriptions.

No commitment. Demo uses your JD to show real screening results.