Projects
Obsidian Agent: Sirimal
A Streamlit-based Obsidian agent designed to help manage and summarize Obsidian notes. It focuses on actionable tasks, follow-ups, and papers to read, making it easier to keep track of work and research using OpenAI's GPT-4o.
Personal LLM Project
Personal project focused on optimizing a Large Language Model to answer about Sri Lanka motor traffic laws. Technologies involved: open source LLMs, PyPDF2, LangChain, text embedding, FAISS as the knowledge base, Streamlit for the interface development.
Profanity Filtering in Speech Content Using Deep Learning
Thesis submission for Master of Business Analytics. The study focused on identifying the F-Words in a speech audio file using audio characteristics such as MFCCs, spectral features, and spectrogram images with the help of deep learning algorithms such as RNN and CNN.
Customer Propensity Model for Seat Booking at Scoot Airlines
Focused on identifying the likelihood of customers buying a seat when booking a trip. Technologies involved: data mining, machine learning, XGB, Python, and statistics, APO deployment.
Enterprise Customer Churn Prediction in Telecommunication Domain
Focused on identifying the likelihood of customers discontinuing service with the company. The project consisted of 2 main phases: customer churn model and churn dashboard development.
Underwater Visible Light Communication Under Turbulent Conditions
Final year project for the BSc in Electrical and Electronic Engineering. The study focused on developing an underwater wireless communication link using visible light as the medium. The system consisted of a transmitter, a receiver, and a demodulation methodology.