Back to Projects

Smart Hire AI

AI-powered recruitment platform with active enterprise pilots. Best FYP Award — UMT Software Engineering Department 2026.

Live Demo
Smart Hire AI screenshot 1

Recognition & Adoption

Best FYP Award

UMT SE Dept. 2026

UMT HR

Platform adoption requested

Enterprise Pilots

Active deployments

Project Overview

Smart Hire AI is a production-grade AI recruitment platform that automates the full hiring lifecycle — sourcing through WhatsApp and voice agents, AI-led screening interviews, rubric-based candidate scoring, semantic candidate search, and team-collaborative analytics. Founded and technically led by me, with active enterprise pilots and adoption interest from the UMT HR department. Also submitted as my BS Software Engineering thesis at the University of Management and Technology, winning the Best FYP Award in 2026.

Key Features

  • Polyglot microservices: Next.js 15 web app + two FastAPI services (business domain + AI/RAG)
  • Natural-language candidate search via LangGraph, Cohere 1024-d embeddings, DuckDB vector index, and Groq Llama-3.x re-ranking
  • Hybrid local/cloud LLM routing with confidence-gated fallback, circuit breakers, and query caching
  • AI voice interview system on Vapi with real-time transcription and MediaPipe/TF.js video proctoring
  • WhatsApp Business (Meta Cloud API) + voice outreach with multi-language detection and fake-candidate scoring
  • Three rubric evaluation modes with mandatory criteria, sub-criteria, and pass thresholds
  • Auto-calculated time-savings ROI of 40+ recruiter hours per campaign
  • ARQ Redis-backed task queue; deployed on Azure VM with Nginx + systemd and GitHub Actions CI/CD

Case Study

Role

Founder & Technical Lead. Lead technical architecture, AI/ML systems, and engineering.

Recognition & traction

  • Best FYP Award — UMT Software Engineering Department 2026
  • UMT HR department requested platform adoption
  • Active enterprise pilots

    • Architecture

    A polyglot microservices system. The Next.js 15 web app handles all user-facing interactions. Two FastAPI services run the backend: a business-domain service for core CRUD and orchestration, and a dedicated AI/RAG service for embedding generation, semantic search, voice interview coordination, and LLM workflows. Background processing runs on an ARQ task queue backed by Redis. Production infrastructure runs on Azure VM with Nginx and systemd, with Railway Cloud for additional services. CI/CD via GitHub Actions.

    Key technical features

  • Natural-language candidate search: LangGraph workflows route queries through Cohere 1024-dimensional embeddings and a DuckDB vector index, with Groq Llama-3.x re-ranking for precision. Hybrid local/cloud LLM routing with confidence-gated fallback keeps latency stable under load.
  • AI voice interview system on Vapi: Live screening interviews with real-time transcription, MediaPipe/TF.js video proctoring, and transcript-based scoring against rubric criteria.
  • WhatsApp + voice outreach pipeline: Multi-language detection, suspicion/fake-candidate scoring, opt-out handling, and seamless handoff between channels.
  • Three rubric evaluation modes: Mandatory criteria, sub-criteria, and pass thresholds with auto-calculated ROI showing 40+ recruiter hours saved per campaign.
  • Resilience patterns: Circuit breakers, query caching, idempotent task processing, and graceful degradation when LLM providers fail.

    • Related Projects

    Vibe Terminal

    Vibe Terminal

    Cross-platform AI terminal assistant — execute shell commands from natural language, with voice I/O and multi-step agent workflows.

    Smart Grade AI

    Smart Grade AI

    AI-powered LMS with RAG-based mentoring — Microsoft Imagine Cup 2025 Global Semifinalist, backed by $25K Azure credits from Microsoft for Startups, in institutional rollout at UMT.

    Josh AI — Recruitment Browser Extension

    Josh AI — Recruitment Browser Extension

    Manifest V3 Chrome extension for sourcing and scoring candidates on Naukri Resdex — India market joint venture.