About

Hi! I’m an AI researcher and engineer with a background in computer engineering and a strong focus on applied machine learning and computer vision.

My work includes research and experimentation with synthetic data generation using GANs, transformers, and diffusion models, as well as real-time violent activity detection using YOLO and ResNet-based architectures. I'm interested in building AI systems that are practical and dependable, with an emphasis on data-driven modeling and applied machine learning, and I’m currently preparing for graduate studies in Computer Science.

Publications

Synthetic Data Generation of Electronic Health Records using CTGAN, Transformers and Diffusion Models

Girban Adhikari, Jivan Acharya, Arjan Sapkota, Subarna Ghimire, Umesh Kanta Ghimire

IOEGC 2025

TLDR

We present a systematic study on synthetic data generation for machine learning, exploring multiple generative approaches including GANs, transformers, and diffusion models and evaluating their effectiveness for downstream tasks. Our work analyzes data quality, distributional similarity, and model performance when trained on synthetic versus real data, highlighting strengths, limitations, and practical considerations for using synthetic data in real-world ML pipelines.

PDF

GitHub

Software

adgirban/ai-doctor-voicebot

Tool

AI Doctor with voice and image recognition, also with Docker implementation.

Python

adgirban/apnr

Tool

Automatic number plate recognition (ANPR) system using deep learning and object tracking.

Python

adgirban/Emergency-Response

Tool

Real-time violence detection system with fine-tuned classification and object detection models.

Python

adgirban/Ghadi-app

Application

Mobile app to customize and visualize 3D watches with Vuforia.

Unity

adgirban/Neoscapes

Game

Neoscapes is a semi-third-dimensional endless runner game with obstacles.

Unreal Engine