Leki Dorji

Portfolio

Projects

Data science and AI work across computer vision, applied machine learning, and practical model development.

01

Facial Emotion Recognition

A deep-learning system that classifies human facial expressions on the FER2013 dataset, comparing a convolutional baseline against a Vision Transformer (ViT).

  • Python
  • PyTorch
  • Vision Transformer
  • CNN
  • Computer Vision
02

Consumer Communication Verification

A machine-learning and NLP-based verification system for consumer emails and SMS messages; using probability scores to classify communications as legitimate, suspicious, or fraudulent and support safer digital interactions at scale.

  • Python
  • NLP
  • Machine Learning
  • scikit-learn
  • Pandas
03

Bird Species Classification

A MATLAB study on fine-grained image classification using the Caltech-UCSD Birds-200-2011 (CUB-200-2011) dataset. The project compares classical computer-vision pipelines (hand-crafted features + classical classifiers) against deep convolutional neural networks, across two image settings — whole images and bounding-box crops — on both a 20-class subset and the full 200-class dataset.

  • MATLAB
  • Computer Vision
  • CNN
  • Image Classification
  • Feature Engineering