- New articles
Two guides to the Business Central API
How to register an app in Azure and Microsoft Entra for OAuth2, then how to prove the connection works in Postman before building anything on it.
The first guide covers the authentication groundwork from start to finish: registering the application in Azure, granting it the right permissions, and wiring it up so Business Central accepts its requests.
The second picks up from there. Before writing any integration code, it shows how to prove in Postman that the Business Central API connection actually works.
Both come out of day-to-day work at Softserve Solution: connecting outside systems to Business Central over REST APIs, with every endpoint validated in Postman before it ships.
Two guides to the Business Central API
How to register an app in Azure and Microsoft Entra for OAuth2, then how to prove the connection works in Postman before building anything on it.
The first guide covers the authentication groundwork from start to finish: registering the application in Azure, granting it the right permissions, and wiring it up so Business Central accepts its requests.
The second picks up from there. Before writing any integration code, it shows how to prove in Postman that the Business Central API connection actually works.
Both come out of day-to-day work at Softserve Solution: connecting outside systems to Business Central over REST APIs, with every endpoint validated in Postman before it ships.
SQL Server backups and Flutter on Windows
Two practical write-ups: scheduling dependable SQL Server backups without any maintenance code, and a straight path to a working Flutter setup on Windows.
The backups guide schedules dependable SQL Server backups with SQL Server Agent and the maintenance plan tooling in SSMS, with no T-SQL to write or maintain. It grew out of automating the backups of a production data warehouse.
The Flutter guide is the straight path through setting up a new development environment on Windows, from downloading the SDK to a working setup.
Running a large language model locally with Ollama
A guide to getting a large language model running on local hardware, and why keeping the data there can matter.
It walks through installing Ollama, pulling a model and talking to it, with no data leaving the machine.
It follows an earlier benchmark of Ollama models, which measured the trade-off between model size and answer quality on ordinary hardware.
Paper published at ICTAR 2024
A custom CNN for lung cancer detection, 98.55% accurate, presented and published in the AI and Data Science Synergy track.
The paper, written with G. R. Perera, describes a custom convolutional neural network that classifies lung CT scans as benign, malignant or normal, built to be smaller and faster to train than established pre-trained models.
It reached 98.55% accuracy, against 97.67% for ResNet50 and 66.15% for VGG16. The data came from Kaggle and Teaching Hospital Anuradhapura, balanced with SMOTE and augmentation, and the results were judged by cross-validation on precision, recall and F1.
The work began as the final year research project, and was presented at ICTAR 2024 with the conference certificate awarded.
Graduated in Software Engineering
Completed a BSc (Hons) in Software Engineering from the University of Plymouth, with Second Class Honours, Upper Division.
Studied at NSBM Green University in Sri Lanka for the University of Plymouth degree, from September 2021 to December 2024.
The final year research, a custom convolutional neural network for lung cancer detection from CT scans, was later published at the ICTAR conference.