Projects
A selection of software engineering projects, web applications, and machine learning models.
Multimodal Deepfake Detector
Engineered a multimodal deepfake detector pairing an Xception visual branch and a CNN audio branch, fused via a cross-modal attention gate and trained on FakeAVCeleb plus 4,000 VoxCeleb2 videos. Achieved 94% accuracy and 0.981 ROC-AUC.
Study Assistant — AI Flashcards & Quizzes
Architected an interactive learning app in Next.js (App Router) supporting 3 study modes: 3D Flashcards, Self-Retrieval Quizzes, and Concept Checklists. Engineered a 5-tier resilient AI fallback backend (Groq, Gemini, NVIDIA NIM).
InterviewIQ — Adaptive AI Technical Interview Platform
Built an adaptive AI technical interview platform that parses resume and job-description PDFs, cross-examines claimed skills against JD requirements, and adjusts question difficulty in real time across Google, Amazon, and Startup interviewer personas. Features JWT auth, Groq/Gemini AI fallback, interview memory for weak-point retesting, a truthfulness auditor with skill-confidence reports, and multimodal voice interaction via the Web Speech API.
Gestara AI — Indian Sign Language Interpreter
Developed an Indian Sign Language interpreter featuring real-time Sign-to-Text and Text-to-Sign translation powered by MediaPipe hand landmarkers, TensorFlow, and a Flask backend with gesture autocomplete.
Devnovate — MERN Stack Blogging Platform
Collaborated in a 3-member team to build Devnovate, a MERN stack blogging platform supporting full CRUD operations, Markdown editing, SEO slug routing, user KPI analytics, dual-database engine, and admin moderation.