sandaru Jayathilaka

I’m an AI/ML Engineer focused on Agentic AI, MCP Servers, and AI system architectures—building scalable machine learning systems and intelligent software that can actually ship. I care about pushing next-generation AI in ways that extend human capability, not just demos.

I earned a B.S. (Hons) in Software Engineering from the University of Plymouth (Dec. 2023) and built a foundation through focused coursework across AI/ML, math/data science (Python), algorithms, and application development (C#, Java). I also continued strengthening my practical depth with certifications like Intro to TensorFlow for Deep Learning, Open Source Models with Hugging Face, and AWS Beginner to Intermediate.

My experience spans computer vision, LLM systems, and production-oriented pipelines. At Sri Lanka Telecom, I worked on an intelligent parking management system using computer vision on NVIDIA Jetson Nano (Python, OCR, YOLO, Linux). At NewNop SL and Kerner Norland, I delivered projects like CNN floor plan segmentation, vehicle monitoring with object detection/tracking + OCR, an in-house Hyperclova AI Manager connecting multiple LLMs, and agentic automation for onboarding—using RAG + agentic workflows to minimize human involvement. Currently, I’m developing MCP-powered, RAG-based fintech and voice solutions, including an MCP Chat assistant for document-driven bookkeeping workflows and an AI customer care voice agent with ASR fine-tuning and TTS training.

I’m especially energized by building reliable, efficient AI systems: integrating LLMs with tools, retrieving context with vector search/RAG, and fine-tuning models for real scenarios. I’m also motivated by turning research into usable products—through pipelines, practical deployments, and clear, user-friendly interfaces.