Research Scientist · Statistical Genomics · AI for Healthcare · Digital Health
PhD | CSIRO Australian e-Health Research Centre | Melbourne, Australia

I am a computational scientist and researcher passionate about unlocking the power of genomics and other health data to understand and prevent human disease. My work sits at the intersection of statistical genetics, AI/machine learning, and digital health infrastructure — building tools and methods that translate genomic insights into real-world clinical impact.

Currently at the CSIRO Australian e-Health Research Centre (AEHRC) in Melbourne, I develop cloud-native genomic platforms, privacy-preserving analytics methods, and scalable bioinformatics pipelines for population-scale datasets.

"The genome is one of the most complex and beautiful datasets in existence — I build the methods and infrastructure to make it speak."

Genomic data is uniquely sensitive — it is immutable, inherited, and identifies not just individuals but entire families. Responsible genomic research demands infrastructure that is ethical by design. I architect and build frameworks that embed data sovereignty, dynamic consent, and access control directly into genomic data platforms.

Key contributions include:

• GeneGuardian — A cloud-native dynamic consent and genomic information management platform, deployed to support Australia's first newborn genomic screening trial (TRAIL). Built on AWS (Cognito · Lambda · DynamoDB · API Gateway · SES) with a ReactJS/NodeJS frontend and Terraform IaC. Implements self-sovereign identity principles, giving participants real-time control over how their genomic data is used.
• GA4GH Standards Alignment — Active involvement in the Global Alliance for Genomics and Health (GA4GH) consortiums, including alignment with Beacon v2 protocols, data access frameworks, and newborn screening genomic standards.
• Privacy-Preserving Analytics — Research into federated learning and secure multiparty computation approaches that allow genomic analyses to be conducted across institutions without raw data leaving its source — a critical requirement for sensitive clinical genomic data.

Stack & Standards: AWS · Terraform · ReactJS · GA4GH Beacon v2 · Federated Learning · SOC2 · FHIR · HL7

The scale and complexity of modern omics data — genomic, transcriptomic, proteomic, and clinical — demands AI systems that go beyond static pipelines. I am actively developing and exploring agentic AI frameworks that can autonomously orchestrate multi-step genomic workflows, reason over heterogeneous data types, and interact with genomic data infrastructure through structured APIs.

Agentic Genomics Infrastructure:

• Designing AI agent systems (using frameworks such as CrewAI and Claude-based agents) to automate tasks like variant QC, cohort curation, phenotype harmonisation, and report generation
• Exploring "agentification" of GA4GH Beacon — replacing static client-initiated query APIs with stateful, reasoning agents capable of multi-hop data discovery across federated genomic nodes
• Building consent governance agents that dynamically interpret participant preferences and enforce access policies in real time

ML/AI for Omics:

• Random Forest & Gradient Boosting for genomic feature selection and variant prioritisation
• Deep Learning (CNN, Transformer-based models) for sequence-level variant effect prediction and multi-omics integration
• Federated & Privacy-Aware ML — training models on distributed genomic datasets without centralising sensitive data
• Multimodal Data Integration

Most large-scale GWAS have been conducted predominantly in European-ancestry cohorts, limiting the generalisability of genetic findings and risk models to diverse populations. My research addresses this gap by conducting large-scale heritability and association analyses across trans-ethnic populations, including extensive collaboration with researchers in Taiwan on Taiwanese biobank datasets.

This work explores how linkage disequilibrium (LD) structure, allele frequency differences, and gene-environment interactions vary across ancestries — and how these differences affect the transferability of PRS models. Improving cross-population genomic inference is essential for equitable precision medicine.

Methods & Tools: Cross-population LD analysis · Ancestry-stratified GWAS · Meta-analysis (METAL) · Admixture modelling · Propensity Score Matching (PSM) · PLINK · REGENIE

Tech Lead | CSIRO × NSW Health Pathology | TRAIL Newborn Genomic Screening Study

A cloud-native dynamic consent and genomic information management platform supporting Australia's first newborn genomic screening trial.

Stack: AWS (Cognito · API Gateway · Lambda · DynamoDB · SES) · ReactJS · MUI · NodeJS · Terraform (IaC)

Key contributions:

• Designed and deployed the full-stack platform from architecture to production
• Implemented self-sovereign identity principles for participant-controlled genomic data access
• Integrated future-proof encryption aligned with GA4GH genomic data standards
• Supported deployment under NSW e-Health Pathology cloud infrastructure

Large-scale genome-wide association studies on biobank-sized datasets (UK Biobank, Taiwanese cohorts) using HPC systems.

Tools: PLINK · REGENIE · BOLT-LMM · GCTA · R · Python · Bash
Methods: Propensity Score Matching, PCA, SNP annotation, epistasis detection, partitioned LD score regression

• Statistical & Deep Learning in Genomics
• Agentic AI in Healthcare and Bioinformatics
• Genetic Data Privacy & Federated Analytics
• Precision & Preventive Medicine
• Digital Health Infrastructure & Policy
• Trans-ethnic Population Genetics
• Science Outreach & Mentorship

Open to collaborations in genomics, digital health, and AI for precision medicine.