Projects
Things I've built — from AI platforms to ML pipelines.
Shortly — URL Shortener with Real-Time Analytics
Production-grade URL shortening service built to handle high-traffic scenarios at scale. Features real-time analytics, AI-generated slugs, Redis caching, async click processing via BullMQ, and a live WebSocket dashboard.
InterviewPrepApp
AI-powered mock interview platform with real-time voice interaction, adaptive difficulty scaling, and structured feedback. Supports both individual practice sessions and mass interview campaigns.
LUMS Navigator
A full-stack campus super-app for LUMS students — centralized platform for societies, eateries, course swaps, community polls, budget tracking, and an AI chatbot trained on the LUMS Student Handbook using a RAG pipeline.
Airbnb NYC Price Prediction
End-to-end ML pipeline predicting Airbnb listing prices using LightGBM. Feature engineering on 48,000+ listings with geospatial clustering and cross-validation, achieving R² of 55.3% and RMSE of $102.93.
Roman Urdu Toxicity Filter
A production-ready content moderation system for Roman Urdu text — fine-tuned mBERT for multilingual toxicity classification, hybrid word-level attribution to identify toxic words without false positives, three-tier moderation actions, and Google Gemini LLM integration with output safety verification.
Brain Tumor Segmentation & Classification
Deep learning pipeline for automated brain tumor detection from MRI scans — U-Net with MobileNetV2 encoder for pixel-level tumor segmentation, EfficientNetB1 for classifying tumors into glioma, meningioma, and pituitary categories, with a Streamlit app for non-technical clinical use.