AI-first software engineer building practical systems.
Strong with full-stack delivery, cloud, and automation.
Focused on shipping user-facing products, not demos.
I'm an AI-focused Software Engineer with 1+ years of industry experience building LLM-integrated systems, RAG pipelines, and agentic AI workflows. Currently pursuing a BSc (Hons) in Information Technology at the University of Moratuwa (CGPA 3.72/4.00).
As an Associate Software Engineer — AI at Bayseian (UK), I architect production AI systems — from a trilingual PII redaction layer (Sinhala, Singlish, code-mixed) using fine-tuned XLM-RoBERTa, to RAG pipelines on Supabase pgvector and n8n agentic workflows wired through MCP servers.
I'm passionate about transformer-based models, vector databases, and multi-agent orchestration. I publish open source — pyvalidx on PyPI and form-validation-lib on npm — and write on Medium about AI/ML engineering and low-resource NLP.
LLMs, RAG pipelines, agentic workflows, fine-tuning
AWS Bedrock, Lambda, EC2, DynamoDB, Cognito
Next.js, FastAPI, ASP.NET Core, scalable apps
Supabase pgvector, MongoDB, MSSQL, Redis
Docker, GitHub Actions, CI/CD, n8n, MCP servers
pyvalidx (PyPI), form-validation-lib (npm)
Python
C#
TypeScript
JavaScript
Next.js
React
FastAPI
ASP.NET
AWS
Azure
LangChainLangGraph
Hugging Face
Supabase
PostgreSQL
MongoDB
Docker
GitHub Actions
n8n
StreamlitBuilding LLM-integrated systems, RAG pipelines, and agentic AI workflows for production multi-tenant deployments.
Developed enterprise HR system features using C# and ASP.NET Core Web API within a microservices architecture; delivered RESTful endpoints and multi-layer business logic.
Honours degree with focus on AI systems, cloud computing, and scalable full-stack applications.
Completed Advanced Level studies in the Bio Science stream before transitioning into software engineering and AI-focused development.
Undergraduate Research · NLP · Low-Resource Languages
RootCode
University of Peradeniya
SLIIT
University Exhibition
Multimodal adaptive learning platform supporting visual, voice, and textual modes with AI-generated content tailored to user learning style. Q&A reasoning engine with chain-of-thought prompting and K-means threshold filtering on the RAG pipeline to distinguish grounded responses from hallucinations.
Open-source Python library for Sinhala subtitle generation from multilingual media. Fine-tuned Meta's NLLB-200 on a curated Sinhala parallel corpus, with timestamp-aware translation, SRT/VTT generation, and INT8 quantized batched inference optimised for 4GB consumer GPUs.
Full-stack hospital management platform; led admin role implementation and built a real-time analytics dashboard for staff and patient data using SignalR. CI/CD on GitHub Actions and Azure Web Services for zero-downtime deployments.
NLP-driven visual generator that converts natural language prompts into hand-drawn-style diagrams using Rough.js and multi-model LLM routing via OpenRouter.
Led a team to design and build an indoor air quality monitoring system using ESP32 and multi-gas sensors (CO2, VOC, PM2.5) with Firebase real-time visualisation and threshold alerting.
Lightweight Python validation library with declarative, zero-boilerplate API for validating user inputs, API payloads, and form data. Published on PyPI.
Declarative JavaScript form validation library published to npm; zero-dependency, with chainable validation rules and custom error rendering for web form inputs.
Feature-complete MERN todo app with authentication, authorization, Redux state handling, Redis caching, and password reset via NodeMailer.
AI-powered PDF extraction and analysis tool for summarization and Q&A built with Flask and Google Generative AI.
Responsive personal portfolio with smooth transitions, expressive layouts, and project-focused storytelling.
Sharing knowledge through articles on Medium
A comprehensive guide to implementing real-time web functionality using SignalR, covering hubs, connections, and scaling strategies.
Learn how to set up automated deployment pipelines using GitHub Actions for .NET applications hosted on Azure Web Services.
Deep dive into building Retrieval-Augmented Generation systems using LangChain, vector databases, and LLM integrations.
How I fine-tuned Meta's NLLB model for Sinhala subtitle generation, addressing challenges in low-resource language NLP.
Have a project in mind or want to collaborate? Let's connect.
Currently open to freelance projects, full-time opportunities, and interesting collaborations in AI and web development.
echo"Feel free to reach out for collaborations or just a friendly chat about tech!"