Hello, I'm Ravindu!

AI Engineer · Small Language Models · MLOps · Robotics

I build AI systems designed for the real world — fine-tuning small language models, shipping end-to-end MLOps pipelines, and exploring where AI meets physical systems.

Previously Research & AI Intern at WSO2. Currently finishing my B.Sc. in Artificial Intelligence at the University of Moratuwa.

Featured Projects

• Adaptive Multi-Agent RAG Platform – Enterprise Knowledge Retrieval

  10 May, 2026

  An in-progress production-oriented RAG system built around LangGraph agent orchestration, pgvector semantic search, Redis caching, and a full-stack React + FastAPI interface — designed to demonstrate how enterprise AI retrieval should actually be built.

• Type II Diabetes Risk Prediction – Published Research

  1 Feb, 2026

  A multiclass ML system predicting Type II diabetes risk from NHANES lab and lifestyle data, published in IEEE. Built a full pipeline with LightGBM, Optuna, MLflow, and feature ablation studies.

• Neuro-Symbolic PC Builder – SLM + Prolog Expert System

  20 Nov, 2025

  A PC configuration assistant that combines a Small Language Model for natural language understanding with a Prolog-based expert system for formal hardware compatibility reasoning — bridged via Janus-SWI.

• Fine-Tune Pipeline – CI/CD-Ready LLM Fine-Tuning Framework

  25 Jun, 2025

  A team-built, production-oriented fine-tuning framework covering LoRA training, inference, and evaluation for Qwen, Llama, and Mistral models — designed to run inside GitHub Actions or Jenkins with full W&B experiment tracking.

• SentiView – AI-Powered Customer Communications Analytics Platform

  18 Jul, 2024

  A full-stack multi-channel sentiment intelligence platform built as a 2nd-year university software project. Spans an Angular 17 dashboard frontend and a FastAPI email analytics microservice with Gemini, AWS Comprehend, MongoDB, and Gmail OAuth.

• AutoChess – Fully Autonomous Chess-Playing Machine

  20 Jul, 2023

  A physical chess board that senses human moves, computes responses with an embedded chess engine, and physically moves pieces using an XY electromagnet gantry — rated at ~2000 ELO.

About

Ravindu Tharuka Weerasinghe — AI undergraduate at the University of Moratuwa (CGPA: 3.86, Dean's List 4/6 semesters), with an IEEE-published research paper and hands-on industry experience.

Previously Research & AI Intern at WSO2, where I fine-tuned small language models with LoRA and co-built an automated MLOps pipeline that runs end-to-end from a GitHub commit to a deployed model.

Highlights

• IEEE Publication

  Type II Diabetes Risk Prediction — 95.27% ROC-AUC on 100k+ NHANES records

• WSO2 Internship

  Research & AI Intern — SLM fine-tuning, automated MLOps pipelines, production AI tooling on Choreo

• TechTriathlon 2024

  Winner

• DataXplore 2026

  Winner

• EXMO 2023

  Exhibited AutoChess — a fully autonomous physical chess-playing machine

Now

• Building an adaptive multi-agent RAG platform with LangGraph, pgvector, and Redis
• Exploring neuro-symbolic AI — combining SLMs with Prolog-based expert system reasoning
• Finishing my B.Sc. in Artificial Intelligence at the University of Moratuwa
• Preparing for an AI engineering role with a long-term focus on robotics and embodied AI

Focus Areas

Small Language Models · MLOps · RAG Systems · NLP · Computer Vision · Robotics · Neuro-Symbolic AI