‹ All Projects
Project Thread🏆 KitaHack 2026

TalentLens — AI Talent Intelligence & Multimodal 3D Interview Platform

🐙 View GitHub

Overview

An end-to-end enterprise recruitment ecosystem designed to eliminate unconscious bias using Multimodal AI and 3D WebGL Avatars to evaluate candidates strictly on competency.

Tech Stack

React.js & ViteThree.js (@react-three/fiber)Google Gemini 2.0 FlashFirebase (Firestore & Auth)Adzuna API

Features

  • Blind Candidate Matrix: Automatic masking of candidate names.
  • 3D Multimodal Interview: Real-time WebGL Avatar with lip-syncing synced to AI speech.
  • Competency Radar Charts: Comparative visualization of technical and behavioral metrics.
  • Deep Semantic Parsing: Extraction of skills and projects from PDFs using Gemini Multimodal vision.

Key Decisions

01

Selected @react-three/fiber to manage the 3D Avatar's jaw-morph targets, allowing for native WebGL lip-syncing with minimal CPU overhead.

02

Engineered a "Blind UI" state within React that strictly restricts access to name and gender fields until the shortlisting phase is finalized.

Learnings

→Learned to decouple the 3D rendering loop from the Gemini API call lifecycle to prevent UI stuttering during multimodal video processing.
→Refined the use of radar charts for normalizing disparate competency scores into a single, judgeable visual metric.
EmailGitHubLinkedInResume