Himeth Wimalagunaratne

// Hello World ๐Ÿ‘‹

Himeth Udana

|

Third-year Computer Engineering undergraduate at the University of Sri Jayewardenepura. I build intelligent systems at the intersection of Agentic AI, Machine Learning, and Full-Stack Development.

Sri LankaOpen to opportunities
scroll

// tech_stack

Tools & Technologies

Technologies I use to build intelligent, scalable systems.

Python
Python
Next.js
Next.js
TypeScript
TypeScript
AWS
AWS
GCP
GCP
Docker
Docker
LangChain
LangChain
PyTorch
PyTorch
PostgreSQL
PostgreSQL
Pinecone
Pinecone

// featured_projects

Projects

A selection of projects focused on Agentic AI, ML, and full-stack engineering.

AllAgentic AIMachine LearningFull-Stack

LegalLens

Agentic AI

AI-driven compliance auditing platform that evaluates documents against Sri Lankan (CBSL, FTRA, LFC) and international (FATF) financial regulations. Uses Agentic RAG to retrieve exact legal clauses and reason over them, producing risk scores with full citation trails.

Next.js
TypeScript
LangChain
Groq
Pinecone
Agentic RAG

Aura

Agentic AI

Multimodal RAG platform that indexes educational videos by synchronizing audio transcripts (Groq Whisper) and visual frame descriptions into a unified vector space. Chat with video content using natural language and get answers with exact clickable timestamps.

Next.js
TypeScript
Python
FastAPI
LangChain
Pinecone
Docker
AWS

LexGuard

Full-Stack

Blockchain-based digital notary dApp that provides immutable proof of existence for legal documents. Leverages Cardano's metadata architecture to store SHA-256 cryptographic fingerprints on-chain, eliminating the need for a third-party notary.

Next.js
TypeScript
Cardano
Blockchain
SHA-256

MalVec

Machine Learning

Multi-class malware family classifier achieving 84% accuracy using opcode frequency analysis and ensemble learning. Combines Random Forest and XGBoost models with Optuna hyperparameter optimization across 21 malware families.

Python
XGBoost
scikit-learn
Optuna
Machine Learning

Traffic-Simulation

Machine Learning

Intelligent traffic management system using reinforcement learning (Q-learning) and computer vision (YOLOv5) to optimize signal timings at intersections. Features a real-time analytics dashboard and adaptive signal control.

Python
YOLOv5
OpenCV
Q-Learning
Reinforcement Learning

Data_Crunch_109

Machine Learning

Competitive data science project applying advanced machine learning techniques for feature engineering, model stacking, and predictive analytics. Demonstrates proficiency in exploratory data analysis and model evaluation.

Python
Jupyter
pandas
scikit-learn
Data Science