AI & Human Oversight Notice
REQUVIA deploys artificial intelligence and large language models (LLMs) strictly as an advisory productivity assistant for human talent acquisition specialists. This notice clarifies our deployment principles, operational boundaries, and oversight safeguards.
Important Non-Guarantee of Hiring Decisions:
REQUVIA AI does not autonomously make hiring, candidate disqualification, or employment offer decisions. All AI outputs are advisory recommendations and drafts that require mandatory human review. Output may be imperfect, and REQUVIA makes no warranty of guaranteed hiring accuracy.
1. Where REQUVIA Uses Artificial Intelligence
Drafting follow-up suggestions for candidates across WhatsApp and Email, explaining application steps, and translating recruiter intent into polite responses.
Summarizing resume highlights against role specifications to help recruiters quickly navigate large talent pools.
Extracting text from uploaded screenshots via PaddleOCR to confirm Workday application form submission status.
Classifying inbound candidate queries (e.g. asking for application link vs rescheduling interview) into structured recruiter queues.
2. Human Oversight & Recruiter Responsibility
- Recruiter Verification: Every message draft generated by REQUVIA AI must be reviewed or confirmed by a human specialist prior to candidate delivery.
- Correction of Hallucinations: Large language models may occasionally generate inaccurate details or misinterpret complex resumes. Recruiters are trained to cross-verify all key data points.
- Bias Mitigation: REQUVIA prompts are engineered to evaluate job-relevant qualifications only, avoiding protected characteristics such as age, gender, race, religion, or disability status.
3. Model Architecture & Data Isolation
Local On-Premises Foundation: REQUVIA's conversational candidate intelligence is powered on-premises by self-hosted Ollama (llama3.2:3b), and optical character recognition for candidate screenshot verification is performed offline by PaddleOCR. Candidate messages, interview transcripts, and uploaded screenshot data remain strictly within the customer's private host environment and are never transmitted to external cloud LLM providers.
Optional Cloud Accelerators: Where an organization explicitly supplies a cloud API key (such as Groq for job description parsing or LinkedIn drafting assistance), prompts are transmitted over encrypted TLS channels under strict commercial terms prohibiting public model training.
No Public Model Training: Client candidate data is never utilized to train public foundation models. Google Gemini is not utilized in the REQUVIA runtime.
4. Questions & Feedback
If you have questions regarding our AI implementation or wish to report an inaccurate recommendation, please submit a report to Grievance Redressal or email hello@requvia.com.