Based in Auckland CBD, New Zealand

Malitha Gunawardhana

Machine Learning, Data Science & Software Engineering Professional

Machine Learning, Data Science, and Software Engineering professional with 5+ years building and deploying applied AI systems and data solutions across medical imaging, sensor networks and production computer vision, plus a PhD in deep learning completed in August 2026. I work end to end, from data pipelines and curation through model training, evaluation, deployment and monitoring, in Python, PyTorch, SQL and cloud/MLOps tooling. I have a consistent record of measurable production impact, including a 70% cut in manual processing time, and I am comfortable translating ambiguous, cross-disciplinary requirements into systems that non-technical users can rely on.

My PhD at the University of Auckland covered deep learning, computer vision and data science. I designed and ran controlled experiments on semantic segmentation, self-supervised learning and network calibration across multiple large imaging datasets, and owned the full data workflow from curation and preprocessing through exploratory analysis, label quality assurance, statistical evaluation and failure analysis. I also built a RAG-based literature agent over a local corpus of research papers using PDF parsing, chunking, embeddings, vector search and an LLM agent to answer natural-language queries with cited source passages.

During my free time, I enjoy tramping, travelling, going to the gym, volunteering, public speaking, and serving as a compere at events.

Email CV LinkedIn Google Scholar GitHub

Current focus Applied AI, machine learning, LLMs and RAG, MLOps, sensor analytics, computer vision, and production software systems

Career

Experience

Industry and research experience across New Zealand, Poland, Germany, the United Arab Emirates, and Sri Lanka.

Machine Learning Engineer (Part-Time)

Viotel Pty Ltd · Parnell, Auckland

March 2026 - September 2026

Built and deployed a deep-learning system for seismic wave phase identification across 3,700+ seismic events and 30,000+ station samples, cutting manual processing time by more than 70%.
Shipped Python time-series anomaly-detection pipelines that flag abnormal vibration patterns across 8+ telecommunications towers, replacing manual inspection with automated monitoring of live sensor feeds.
Ran data-quality and diagnostic analysis on raw sensor streams with geophysicists and a three-member field team, isolating instrument faults, validating model outputs against expert labels, and reporting findings that guided deployment of 17 seismic sensors across the Taupō region.

Time-series analysisAnomaly detectionSignal processingDeep learning

Machine Learning Engineer (remote from Dec. 2023)

Institute of Fundamental Technological Research, Polish Academy of Sciences (IPPT PAN) · Warsaw, Poland

September 2022 - January 2025

Delivered deep-learning models for breast tumour identification from ultrasound images, reaching 86% identification accuracy and enabling a low-cost computer-aided detection system for screening in resource-limited settings.
Designed reusable, reproducible medical-imaging pipelines covering data sourcing, curation, preprocessing, quality assurance, training, evaluation and cross-dataset comparison, so new datasets could be onboarded without rebuilding the stack.
Partnered with clinicians, researchers and engineers to define requirements, agree evaluation criteria, communicate model limitations, and keep technical outputs usable by non-ML stakeholders.

Artificial Intelligence Researcher (concurrent with IPPT PAN)

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) · Abu Dhabi, UAE

September 2022 - September 2023

Developed self-supervised and semi-supervised video-analysis models that held performance under limited labelled-data conditions, reducing the volume of manual annotation required.
Ran controlled experiments on LLMs, network calibration and deep-learning reliability, quantifying failure modes and producing evaluation evidence that informed model-selection decisions.
Led a four-member team through experiment design, analysis and delivery of findings, published and presented at leading artificial intelligence conferences.

Machine Learning Engineer

PromiseQ GmbH · Berlin, Germany

June 2022 - November 2022

Improved deep-learning models behind a live surveillance platform running across 3,000+ CCTV cameras, diagnosing recurring failure modes and prediction errors in production traffic.
Reduced false alarms by 5% through network calibration and error analysis, measurably improving the reliability of a tool used by operators every day.
Worked with cross-functional teams to monitor deployed-model performance, triage production incidents, and iterate on system reliability.

Full-Stack Software Engineer

Xeptagon (Pvt) Ltd · Colombo, Sri Lanka

March 2021 - May 2022

Worked directly with senior stakeholders, including CEOs, CTOs and managers, to understand operational processes, clarify ambiguous and evolving requirements, and translate them into shipped production systems.
Led a four-member team delivering a low-latency domain drop-catching system with over 90% success at under 50 ms latency, integrating concurrent backend components and optimising execution performance.
Built audio-processing functionality for a student learning management system, transforming 80+ audio signals into reusable structured features through signal processing and feature engineering.

Software Engineer (Intern)

Synergen Technology Labs (Pvt) Ltd · Colombo, Sri Lanka

June 2019 - December 2019

Designed and supported development of a wearable physiological-signal acquisition device, and collected an original dataset through controlled stress-inducing experiments.
Built physiological signal-processing and ML pipelines for stress classification, achieving over 90% accuracy across four stress categories.

Academic background

Education

Doctor of Philosophy (Deep Learning, Computer Vision & Data Science)

Dec. 2023 - Aug. 2026

University of Auckland · Auckland, New Zealand

Thesis: Robust, Uncertainty-Aware, and Scalable Deep Learning for Bi-Atrial Segmentation from LGE-MRI in Atrial Fibrillation. Award: Health Research Council Scholarship.

Designed and ran controlled experiments on semantic segmentation, self-supervised learning and network calibration across multiple large imaging datasets, applying cross-validation, statistical hypothesis testing, uncertainty quantification and error analysis to compare competing approaches and communicate results to clinical and engineering stakeholders.

Owned the full data workflow: data curation, cleaning, preprocessing, exploratory data analysis and label quality assurance, resolving distribution shift and annotation inconsistency across sources.

Built reproducible Python ML pipelines with experiment tracking and version control, enabling systematic benchmarking, cross-dataset validation and failure analysis.

Built a RAG-based literature agent over a local corpus of research papers using PDF parsing, chunking, embeddings and vector search with an LLM agent, answering natural-language queries with cited source passages and cutting literature-search time.

B.Sc. Engineering Honours

Jan. 2017 - July 2021

University of Moratuwa · Moratuwa, Sri Lanka

Dean's List placement in Semester 7. Key modules: Statistics, Image Processing and Machine Vision, Neural Networks and Fuzzy Logic, Medical Imaging, and Signal Processing.

Capabilities

Technical Skills

Programming Languages

Python (Pandas, NumPy), SQL, Bash, TypeScript, MATLAB

Machine Learning

PyTorch, TensorFlow, Keras, scikit-learn, XGBoost, computer vision, deep learning, time-series modelling

Statistics & Experimentation

Hypothesis testing, cross-validation, uncertainty quantification, model calibration, experiment design, error analysis

Generative AI & LLMs

RAG, LangChain, CrewAI, Weaviate, LLM agents, agentic workflows, prompt engineering

Data & MLOps

MLflow, Apache Airflow, PySpark, Weights & Biases, Neptune, Docker, Jenkins CI/CD

Cloud & Engineering

AWS, Azure, GCP, Git, Linux, REST APIs, MongoDB, Jira, Confluence

Data Analytics & BI

Power BI, Microsoft Excel, statistical analysis, data visualisation

Certifications & Training

Oracle Cloud Infrastructure 2024 Generative AI Certified Professional · LLM fundamentals, pre-training, fine-tuning and alignment, generative AI, MLOps Microsoft Certified: Azure Fundamentals (AZ-900) · Azure cloud services, storage, networking, security and virtualisation Large Language Model Agents, UC Berkeley · LLM agents, multi-agent coordination, prompt engineering, RAG, MLOps Data Science Career Track, 365 Data Science · Python, SQL, data wrangling, statistical analysis, model evaluation

Research

Selected Publications

Selected work across medical imaging, computer vision, video understanding, model calibration, and signal processing. View the complete publication list on Google Scholar.

2026 · MIDL

A Comprehensive Benchmarking and Systematic Analysis of Deep Learning Models for Sonomammogram Segmentation

Malitha Gunawardhana and Norbert Żołek

Publication

2026 · Medical Image Analysis

MBAS2024: A Large-Scale Benchmark for Multi-Class Bi-Atrial Segmentation in Multi-Center Contrast-Enhanced MRIs

Fangqiang Xu et al., including Malitha Gunawardhana

Publication

2026 · STACOM

TASSNet: A Deep Learning Framework for Robust Bi-Atrial Segmentation for Assessing Structural Basis of Atrial Fibrillation

Malitha Gunawardhana, Mark L. Trew, Gregory B. Sands, and Jichao Zhao

Publication

2025 · Discover Artificial Intelligence

Integrating Deep Learning in Cardiology: A Comprehensive Review of Atrial Fibrillation, Left Atrial Scar Segmentation, and the Frontiers of State-of-the-Art Techniques

Malitha Gunawardhana, Anuradha Kulathilaka, and Jichao Zhao

Publication

2025 · Computers in Biology and Medicine

BSA-Net: Boundary-Prioritized Spatial Adaptive Network for Efficient Left Atrial Segmentation

Fangqiang Xu et al., including Malitha Gunawardhana

Publication

2025 · IEEE EMBC

ResNext-Based U-Net for Segmenting Sonomammogram

Malitha Gunawardhana and Norbert Żołek

Publication

2025 · DL-HAR / IJCAI

How Effective Are Self-Supervised Models for Contact Identification in Videos?

Malitha Gunawardhana, Limalka Sadith, Liel David, Daniel Harari, and Muhammad Haris Khan

Publication

2025 · arXiv

CrossVideoMAE: Self-Supervised Image-Video Representation Learning with Masked Autoencoders

Shihab Aaqil Ahamed, Malitha Gunawardhana, Liel David, Michael Sidorov, Daniel Harari, and Muhammad Haris Khan

Publication

2025 · STACOM-MICCAI

ResNet-Based Convolutional Framework for Segmenting Left Atrial Scars and Cavities

Malitha Gunawardhana, Fangqiang Xu, Yun Gu, and Jichao Zhao

Publication

2024 · CVPR

Towards Generalizing to Unseen Domains with Few Labels

Chamuditha Jayanga Galappaththige, Sanoojan Baliah, Malitha Gunawardhana, and Muhammad Haris Khan

Publication

2024 · CVPR

Why Not Use Your Textbook? Knowledge-Enhanced Procedure Planning of Instructional Videos

Kumaranage Ravindu Yasas Nagasinghe et al., including Malitha Gunawardhana

Publication

2023 · CVPR

Multiclass Confidence and Localization Calibration for Object Detection

Bimsara Pathiraja, Malitha Gunawardhana, and Muhammad Haris Khan

Publication

2023 · ICIIS

Evaluation of Noise Reduction Methods for Sentence Recognition by Sinhala Speaking Listeners

Malitha Gunawardhana, Chathuki Navanjana, Dinithi Fernando, Nipuna Upeksha, and Anjula de Silva

Publication

Beyond engineering

Service & Leadership

Feb. 2024 - Aug. 2026

Student Council Member

Auckland Bioengineering Institute, University of Auckland

Elected Student Council Member representing postgraduate students in university governance, student advocacy, welfare initiatives and policy discussions.

July 2024 - Apr. 2025

Mentor

Sustainable Education Foundation

Mentored students through structured academic, professional and personal development guidance.

July 2022 - July 2024

Vice President for Club Service

Rotaract Club of Alumni of the University of Moratuwa

Coordinated member engagement, fellowship activities and community initiatives.

DBLP ORCID Medium