Nawaraj KhanalAI / ML engineer

Machine learning for practical problems.

I am Nawaraj Khanal, an AI/ML engineer based in Kathmandu. I work on machine learning projects in fraud detection, explainable prediction and computer vision. I am especially interested in graph learning and making model decisions easier to understand.

See my projectsDownload my CV

Interested in applied machine learning and ML engineering opportunities. Main areas: Python, PyTorch, graph learning, computer vision and explainable AI.

Normal accountsCluster flagged for review

Selected work

Transactions can look ordinary on their own. Looking at the links between accounts can reveal patterns that are harder to spot in individual records.

Graph-based fraud and anomaly detection

I represented financial transactions as a graph, used Node2Vec and graph-learning methods to capture relationships between accounts, then combined anomaly detection with supervised classification. SHAP explanations and a Streamlit interface help reviewers inspect flagged activity.

  • PyTorch Geometric
  • NetworkX
  • Neo4j
  • Node2Vec
  • SHAP
  • Streamlit
Read the case study
Result
Anomaly recall improved by 20%
Scale
Experiments on graphs with over 1 million nodes
Recognition
Best System Award

More projects

Loan prediction with explanations

A machine learning model for loan default risk, with SHAP explanations to help inspect the features that influence each prediction.

View the code

Heart disease prediction

A prediction pipeline using Logistic Regression and SVM, with SHAP explanations to show which features influence each result. The reported AUC-ROC is 0.88, and the project received Best Project Proposal in 2024. It is an educational project, not a diagnostic tool.

View the code

Computer vision experiments

Prototypes for drowsiness detection, object tracking and medical image classification using OpenCV, YOLOv8 and PyTorch. The work focuses on testing model behaviour and improving runtime efficiency.

View the code
In progress

ML Systems & MLOps Lab

A work in progress focused on connecting model training with an API, containers, tests and experiment tracking. Completed work is kept separate from planned milestones.

See the plan
All projects in detail

From model development to practical use.

I enjoy working through the full machine learning process: preparing data, choosing suitable metrics, checking where a model fails and building interfaces that make its results easier to use.

My current learning focus is the engineering around machine learning, including APIs, Docker, testing and experiment tracking. I describe these as works in progress unless they have been completed and tested.

More about me and how I work