AI Platform2025

Recruvia

Full-stack AI recruitment platform with semantic resume matching via Sentence-Transformers, voice-based AI interviews powered by Mistral AI and RAG, and real-time proctoring using TensorFlow.js.

Built by Abdullah Nadeem, AI Engineer

View Live Project
Recruvia interface screenshot

Overview

Recruvia is a full-stack AI recruitment platform built as three Next.js portals on Supabase PostgreSQL, with a Python background worker running on Railway.

It automates the hiring funnel end to end: an ATS engine screens resumes semantically, qualifying candidates are invited to a voice-based AI interview, and a proctoring layer checks interview integrity in real time.

What I Built

  • Automated ATS engine using Sentence-Transformers for semantic resume matching, RoBERTa NER for entity extraction, and fuzzy skill matching
  • Automatic interview invitations sent through Resend to candidates who pass screening
  • Voice-based AI interviewer powered by Mistral AI with RAG retrieval from Pinecone, adaptive questioning, and real-time transcription via the Web Speech API
  • Real-time proctoring with TensorFlow.js (BlazeFace, COCO-SSD) monitoring seven violation types, with snapshot-based evidence capture and automated integrity verdicts

Tech Stack

  • Next.js
  • Python
  • Supabase
  • Mistral AI
  • Pinecone
  • TensorFlow.js