Currently developing AI systems at WaveSkill and AiRentoSoft, with expertise spanning Agentic Systems, Deep Learning, Machine Learning, and scalable infrastructure.

My dual-degree background in Computer Science (First Class Honours) and Information Systems (4th Year) provides both technical depth in AI/ML and a business systems perspective. Having published a research paper and an open-source package for the community, I am passionate about transforming complex AI innovations into deployed solutions that deliver measurable real-world impact.

• 🔬 Research Focus: RAG Systems, Agentic AI, AI Automation, ML/DL with Mathematical Optimization, NLP & LLMs in Sinhala Machine Translation
• 🎵 Fun Fact: I'm also a bassist and a professional musician when not coding!

AI/ML Engineer - Freelance @ WaveSkill Pvt. Ltd. (December 2025 - Present)

• Designed and engineered the AI-powered financial document extraction pipeline with computer vision models and helper agents to extract financial data from company statements.
• Developed the fundamental analysis engine for the Stock Market API to give users stats and AI-powered insights on companies regarding their financial positions.
• Developed an Agentic Shopping assistant for Baby Care products e-commerce website (TOGO).
• Developed a TTS dataset and trained a VITS Model for Sinhala.

AI/ML Engineer - Intern @ AiRentoSoft Pvt. Ltd. (March 2026 - September 2026 | 6 months)

• Developed a vehicle damage detection system with DL models, API and Mobile app.
• Collaborated in further developing the AiRentoSoft System which includes the Web app, Mobile app, Reservations Plugin and AI Caller Agent.
• Assisted in business leads extraction for potential clients for improving business opportunities.
• Collaborated with the developers, business support team and QA engineers via Jira with scheduled sprints and meetings.

AI/ML Engineer - Intern @ Residue Solutions Pvt. Ltd. (September 2025 - February 2026 | 6 months)

• Engineered an AI-powered shopping assistant (FashionHub.ai) for customers to get assistance on product search, Virtual Try-On, Cart management, Payment assistance, QnA and Size matching.
• Leader of the AI/ML Intern team of five, providing technical guidance on project architecture, code reviews, and implementation strategies and regular meetings to ensure milestones were met.
• Conducted research and integration by collaborating and leading the team members with LLM orchestration techniques, reasoning and multi-model pipelines that enabled complex product recommendations and Virtual Try-On features.

System Support Officer - IT @ Mercantile Investments FPLC (2022 | 3 months)

• Assisted in hardware, network maintenance and repairs, significantly reducing issues for users through proactive diagnostics and component-level troubleshooting.
• Resolved technical queries in Help Desk support, providing rapid software, OS, and peripheral solutions for diverse end-user requirements.

• 🧠 Tech: YOLO, PyTorch, FastAPI, React, HuggingFace, WandB, Kaggle, MLOps, Roboflow
• 🎯 Developed a Vehicle Damage Detection system consisting of a combination of YOLO models with a FastAPI wrapper hosted in Huggingface spaces.
• 📈 Pre-trained the models on Image datasets, versioned and deployed using MLOps tools.
• 📱 Further developed the live mobile app to support new features which is in Playstore and Appstore.

• 🧠 Tech: React, Node.js, .NET, Retel.ai, OCR, Azure DevOps
• 🎯 Collaborated in Developing and Implementing new features including AI Integrations across web and mobile platforms.
• 💬 Added AI Integrated features like document scanning, card scanning using OCR models in both web and mobile apps.
• 📞 Managed and worked on the AI Caller agent in Retel.ai for client requirements.
• 🛠️ Scaled React web applications, reservations plugins, .NET API, and developed the 2.0 version of the mobile app with documentation in Azure DevOps.

• 🧠 Tech: Playwright, Selenium, Python, ETL
• 🎯 Implemented a business leads extraction pipeline for extracting potential clients in USA based car rental businesses using automation scripts and web scraping.
• 📊 Processed scraped data using ETL pipelines into understandable formats for business support division to use.

• 🧠 Tech: Coqui VITS, PyTorch, Kaggle
• 🎯 Developed a research purpose private TTS model with support for Sinhala language.
• 🎙️ Created a private dataset from scratch for VITS format to train the Speech Model on Kaggle dual T4 GPUs.
• 🚀 Gained research level quality output for product pitching and further funded development.

• 🧠 Tech: FastAPI, Postgres, CrewAI, LangGraph, Langfuse
• 🎯 Developed an Agentic Shopping assistant for e-commerce website for baby care products.
• 💬 Implemented features in Product search, QnA using store policies and automated customer support, live on the TOGO website.

• 🧠 Tech: Next.js, FastAPI, Weaviate, Ollama, LangGraph, LangChain, Kafka, Groq, Gemini, MongoDB, Google Vertex AI
• 🎯 Intelligent retail assistant utilizing a ReAct capable Agent with DAG-based task execution.
• ⚡ High-throughput MongoDB to Weaviate Kafka pipeline built for real-time data synchronization.
• 🤖 Modular multi-LLM factory (Gemini, Llama) and Google Vertex AI for generative Virtual Try-On capabilities.

• 🧠 Tech: FastAPI, LlamaIndex, OpenCV, Groq, MongoDB, LLMs, OCR
• 🎯 Designed and engineered a scalable financial intelligence API using FastAPI for Fundamental analysis of Colombo Stock Exchange (CSE) filings.
• 📊 Built analysis engine calculating core financial ratios and growth metrics, integrating AI-driven insights for performance evaluations.
• 🗃️ Developed extraction engine utilizing hybrid OCR and semantic mapping with LLMs to transform unstructured quarterly statements into standardized datasets.

• 🧠 Tech: FastAPI, MongoDB, React 19, TypeScript, Zustand 5, Tailwind CSS, LangChain, LangGraph, Groq, Docker, Recharts, Pydantic v2
• 🎯 Enterprise-grade, asynchronous reporting system and analytics dashboard with Role-Based Access Control (Contributors, Managers, Admins) and ISO calendar standardization.
• 🤖 Intelligent AI assistant powered by LangGraph and LangChain with Groq LLM featuring a 5-response rolling summarizer and grounded MongoDB RAG.
• 📈 Real-time KPI tracking, compliance rate metrics, and interactive charts built with Recharts.
• 🐳 Containerized with Docker Compose for seamless multi-service orchestration.

Final Year Research Project

• 🧠 Tech: TensorFlow, RNN, Bi-LSTM, Flutter, FastAPI, Celery, Redis, MongoDB, TensorFlow Serving, GCP, MinIO S3
• 🎯 Intelligent system performing musical analysis on audio for musicians, exploring Bi-LSTM networks and time-frequency feature extraction (+80% accuracy).
• 📱 Practical implementation using FastAPI, MongoDB, and TensorFlow Serving with a Flutter-based mobile UI for real-time analysis and playback integration.
• 📦 Published as a PyPI package with CI/CD deployment to Google Cloud Platform.

• 🧠 Tech: Apache Spark 3.5, Apache Kafka, Apache NiFi, MLflow, LightGBM, Prometheus, Grafana, Google Cloud BigQuery, PySpark, Python, Docker, Pandas / PyArrow
• 🎯 Enterprise-grade distributed streaming intelligence pipeline scoring transactions and detecting fraudulent checkout events in real time (< 50ms SLA).
• 🧹 Dropped 339 proprietary V-columns and 38 obscured masks across 590k+ IEEE-CIS records, building a 30-feature reproducible engineering pipeline with cyclical temporal encodings and interaction keys.
• ⚡ Vectorized PyArrow / Pandas UDF inference on PySpark Structured Streaming with singleton LightGBM Booster caching (sub-5ms scoring).
• 💰 Formulated financial cost-matrix optimization ($\tau^* = 0.10$) balancing False Negative chargebacks ($150) against False Positive operational friction ($10).
• 🔄 Dual-sink micro-batch routing dispatching high-risk alerts to Kafka (transactions.alerts) and complete audit logs to Google Cloud BigQuery (fraud_analytics.transactions_log).
• 📊 Full observability stack with Prometheus scraping metrics and a pre-configured Grafana telemetry dashboard tracking throughput, consumer lag, and model drift.

Group Project for USJP

• 🧠 Tech: Django, FastAPI, Next.js, Neo4j, PostgreSQL (Supabase), MongoDB, Redis, Supabase, Azure, GitHub Actions, Azure Bicep
• 💬 Social application for university students featuring media sharing, direct messaging, and forum-based discussions.
• 🏗️ Developed backend in Django and FastAPI with PostgreSQL (Supabase), MongoDB, Neo4J, and Redis, deployed on Azure with GitHub Actions & Azure Bicep.

• 🧠 Tech: CrewAI, LangChain, Ollama, Mistral LLM, ChromaDB, HuggingFace
• 📚 Research assistant leveraging LangChain, CrewAI, and Mistral LLM (via Ollama) for research paper analysis and academic Q&A.
• 🔍 Implemented document ingestion, summarization, topic extraction, and semantic retrieval using HuggingFace embeddings with ChromaDB.

• 🧠 Tech: Streamlit, LangChain, Groq, Qdrant, HuggingFace, MongoDB, Llama LLMs, RAG
• 🎯 Semantic resume-to-job description matching platform using Retrieval-Augmented Generation (RAG).
• 📊 Automatically analyzes resumes, generates job descriptions, and provides ranked job recommendations with similarity scores.

• 🧠 Tech: PyTorch, OpenCV, YOLO, LangChain, Ollama, LLM, Streamlit, Telegram Bot
• 🚨 Real-time camera feed threat detection system combining YOLO computer vision with LLM scene understanding.
• 🤖 Translates suspicious visual detections into detailed descriptions notified via Telegram Bot.

• 🧠 Tech: PyTorch, TensorFlow, OpenCV, U-Net, CNN, Albumentations, Kaggle
• 🏥 Medical-grade U-Net encoder-decoder model built on 2D and 3D MRI image datasets.
• 📊 Achieved ~90% accuracy using DICE coefficient, IoU, and custom loss functions with Albumentations data augmentation.

• 🧠 Tech: Python, SQL, Snowflake, ETL, XGBoost, Streamlit
• 🔗 Real-time ETL pipeline extracting LinkedIn job postings to Snowflake for warehousing and processing.
• 📈 Machine learning regression and classification models (XGBoost) for analyzing job trends, visualized through an interactive Streamlit dashboard.

• 🧠 Tech: Transformer, LLM, Hugging Face, MBart50, BitsAndBytes
• 🇱🇰 Fine-tuned translation model for English-to-Sinhala subtitles.
• ⚡ 8-bit quantized implementation optimizing model weights for efficient local deployment.

• 🧠 Tech: PHP (CodeIgniter), HTML, CSS, JavaScript, MongoDB, Bootstrap
• 🕷️ Web scraping bot collecting real-time e-commerce pricing data.
• 📱 Fully responsive layout delivering cross-platform price comparison.

• 🧠 Tech: MATLAB, Neural Networks
• 📊 Feedforward neural networks built for biometric user verification and pattern recognition.

• 🧠 Tech: .NET, C#, XAML, SQL Server
• 👥 Desktop application managing user bookings, billing, vehicle status, and returns.

• 📝 Research Paper: Published "Musical Chord Recognition using Deep Learning" in the journal at the NSRSIT'25 Conference.
• 📦 Open Source: Published the accordo.ai PyPI package with developed DL Models for community use.

• BSc (Hons) Computer Science - First Class Honours @ University of Plymouth, UK (NSBM Green University) (2022 - 2025)
• BComp (Hons) Information Systems (4th Year) @ University of Sri Jayewardenepura (2023 - Present)
• Dip. in Technology @ Esoft Metro Campus (2019)
• A/L - Math Stream (A1, C2 | O/L: A7, B2) @ St. Peter's College, Colombo - 4 (2008 - 2021)