Adaptive AI USMLE prep platform with fine-tuned LLM, real-time analytics, and AI question banks. Scales to 1K+ users, improving study efficiency ~40%.

Web chess platform with Coach David AI, engine-backed moves, real-time multiplayer rooms, board themes, and plain-language move feedback.

Text-prompt → 3D model pipeline with real-time Three.js browser preview, FastAPI backend, and Blender mesh post-processing. Dockerized deployment.

Fully in-browser ATS resume parser with OCR via Tesseract.js + PDF.js and one-click LaTeX resume generation. Zero backend, custom Aquamarine design system.

Full-stack MERN social app with JWT auth, Redux Toolkit state, and MongoDB social-graph schemas reducing feed-load response by ~35%.

All-in-one image & PDF editor — merge, split, compress, watermark, canvas transforms, and server-side PDF ops via FastAPI. Deployed on Vercel with CI/CD.

Three-module medical CV pipeline detecting 7 fracture types from X-rays. DenseNet-121 at 74.56% accuracy + YOLOv8 detection + SegNet segmentation.

Anime face generation (DCGAN), sketch-to-color via U-Net + PatchGAN (Pix2Pix), and unpaired domain translation (CycleGAN) — trained on Kaggle T4 GPUs.

Full PyTorch transfer learning pipeline for multi-class classification (Human / Animal / Non-Living) with data augmentation and automatic checkpointing.

Real-time webcam phone detection with custom training, mAP evaluation, and multi-format export (ONNX, TorchScript, TFLite) for edge deployment.

Full DS lifecycle predicting exam scores. Best model: R² = 0.721 on final exam prediction. Interactive Streamlit dashboard for at-risk student identification.

Intelligent Sudoku solver with arc-consistency checks (AC-3) — 80% faster than brute-force backtracking on hard-rated puzzles.