I'm the bioinformatician at the ʻIolani School Office of Community Science in Honolulu. I build production Nextflow and Snakemake pipelines for Oxford Nanopore data, and the cloud platform that lets teachers, students, and field researchers run them without touching a terminal.
- Pipelines: bacterial genome assembly, eDNA species ID, and 16S microbiome profiling (Nextflow DSL2, containerized, versioned releases)
- Platform: a GCP web app (React, FastAPI on Cloud Run, Google Batch, Firebase auth) that runs pinned pipeline releases on demand, view the live platform demo
- Research: reference genomes and phylogenomics for Hawaiian biodiversity, with SDZWA, Bishop Museum, the University of Hawaii, and NASA collaborators
- Outreach: training Hawaiʻi DOE teachers through the ʻĀina Informatics Network to bring real sequencing into classrooms
| Pipeline | What it does | Highlights |
|---|---|---|
| edna-ont-nf | Taxon ID from mixed ONT eDNA amplicon pools | Modern replacement for decona: isONclust quality-aware clustering, racon/medaka consensus, BLAST with explicit no-hit/low-identity flagging so undescribed endemic sequences aren't silently dropped |
| 16S-nf | ONT 16S rRNA microbiome profiling | Same architecture as edna-ont-nf, plus Bray-Curtis/PCoA beta diversity; shares an output schema so one frontend visualizes both |
| ulana-nf | Bacterial whole-genome assembly and characterization | Flye → Medaka → Prokka → CheckM → AMRFinderPlus, with per-step toggles. A Nextflow port of ulana-ht (Snakemake) |
| NovoClust | De novo variant discovery and abundance tracking from long-read amplicons | Clustering parameters validated against known copy numbers |
Every pipeline runs locally with Docker, Singularity, or Conda, and on Google Batch in the cloud.
flowchart LR
U["Teacher / student / researcher"] --> F["React frontend<br/>(Firebase auth)"]
F --> B["FastAPI backend<br/>(Cloud Run)"]
B --> N["Nextflow<br/>pinned pipeline release"]
N --> G["Google Batch<br/>containerized tasks"]
G --> S[("Results<br/>tables · plots · FASTA")]
S --> F
- Pinned releases: the platform calls a tagged version of each pipeline, so every result is reproducible and citable.
- No cluster to babysit: Google Batch spins compute up per job and back down afterward.
- Least-privilege by default: CI/CD and runtime use narrowly scoped service accounts.
- pwviz: public dashboard for the Paepae O Waikolu stream-biodiversity program
- ulana-gui, decona-gui, epi2meviz-reboot: R Shiny front ends that made ONT tools usable in classrooms (the predecessors of the platform)
Workflows Nextflow · Snakemake · Docker · Singularity
Cloud Google Batch · Cloud Run · Firebase · GitHub Actions
Languages Python · R · Bash · FastAPI · React · R Shiny
Genomics Oxford Nanopore · hifiasm · Flye · Medaka · Hi-C scaffolding · BUSCO · IQ-TREE · BEAST2
ResearchGate · Publications in Molecular Phylogenetics and Evolution and Astrobiology · Speaker, Oxford Nanopore London Calling 2024