Nov 14, 2025

TwinlyAI

Link to open source: https://github.com/KurianJose7586/HackfestKevlarSubmission.git

Link to Live Project: https://twinly-ai.vercel.app/

What This Build Is About

TwinlyAI is a next-generation recruitment intelligence platform that fundamentally changes how hiring is done. It transforms traditional, static resumes into interactive, conversational AI twins.

Instead of manually scanning hundreds of PDFs, recruiters can now talk directly to a candidate's profile, asking specific questions or even conducting a full, real-time voice interview with the AI.

The platform is built on two core features:

AI Candidate Twins: A user can sign up, log in, and upload their resume. The system uses a custom RAG (Retrieval-Augmented Generation) pipeline (powered by FAISS, BAAI embeddings, and Llama models on Groq) to instantly create a custom chatbot trained exclusively on that resume's data.
The Recruiter Super-Bot: This is an AI assistant for recruiters that allows them to perform two powerful actions:
- Natural-Language Shortlisting: Recruiters can use natural language prompts like, "Show me backend developers with Python + FastAPI experience who’ve worked in fintech," to instantly search the entire database of candidates.
- Live AI Pre-Screening: This is the flagship feature. Recruiters can initiate a live, "Google Meet" style voice call with a candidate's AI twin. Using Agora for real-time Speech-to-Text (STT) and Text-to-Speech (TTS), the AI can hold a natural, conversational pre-screening interview, available 24/7 with zero scheduling.

Why I Built This

Traditional hiring is slow, repetitive, and inefficient. I built TwinlyAI to solve the key pain points I saw in the process:

The Problem: Recruiters are forced to manually scan hundreds of resumes, where valuable skills and context are often buried in text. Great candidates get overlooked. Furthermore, scheduling initial "pre-screening" calls is a logistical bottleneck that can take days.
The Vision: I wanted to build a system where recruiters don't just search for talent—they talk to it. My goal was to prove that resumes shouldn't just be static documents; they should be intelligent, conversational interfaces that can advocate for the candidate.
The Solution: TwinlyAI saves recruiters hours of manual scanning and eliminates the scheduling bottleneck. It gives candidates a way to stand out by showcasing their knowledge and communication style, not just keywords on a page.

How It Can Be Useful for Others

TwinlyAI is designed as an AI engine that can be valuable for multiple groups, not just as a standalone platform.

🔹 For Recruiters & Hiring Managers

Instant Shortlisting: Convert hours of manual resume scanning into a single, natural-language query.
Zero-Schedule Pre-screening: Conduct AI-powered voice interviews with candidates' AI twins 24/7, getting immediate, nuanced answers.
Build a Conversational Talent Pool: Quickly understand a candidate's true experience and fit without ever sending a scheduling link.

🔹 For Candidates

Stand Out: Convert a "boring" PDF resume into an interactive, intelligent AI twin that can't be ignored.
Showcase Your Skills: Let your AI profile do the talking and demonstrate your knowledge and experience in a conversational context.

🔹 For Career Platforms / Job Portals (The "Aha" Moment) We realized we don't need to create another job portal. Recruiters and candidates already meet on platforms like LinkedIn, Internshala, and Naukri.

TwinlyAI is designed to be the AI engine that powers these existing ecosystems. It can integrate via a simple API or plugin to add an "Start AI Voice Interview" button directly to candidate profiles on the platforms recruiters already use. This makes it an integration-ready feature, not a new marketplace that needs to be grown from zero.

🔹 For AI Builders & Hackathon Teams This project serves as a practical, full-stack template for combining several cutting-edge technologies: