Bioinformatics Platform Engineer — Long-Read Sequencing & Structural Variants

Summary

7 years building analysis infrastructure for Oxford Nanopore and PacBio HiFi long-read data. One of the first production users of Dorado basecalling at scale; reduced basecalling cost 55% by porting GPU batching logic to CUDA kernels. Built the SV and tandem repeat analysis platform at Element Biosciences supporting flagship customer datasets. Comfortable operating at the intersection of bioinformatics and systems engineering.

Experience

Senior Bioinformatics Engineer — Element Biosciences (2022–Present)

• Designed and deployed Kubernetes-based long-read analysis platform on AWS; auto-scales from 0 to 500 concurrent Snakemake jobs.
• Built Rust CLI for real-time adaptive sampling decision engine (Read Until API); 30% enrichment efficiency improvement over published baseline.
• Implemented TRGT + Straglr tandem repeat genotyping pipeline for rare disease programme covering 60 known STR loci.

Bioinformatics Scientist — PacBio (2018–2022)

• Co-developed PBSV structural variant caller (C++/Python); shipped in PacBio SMRT Link v10–12.
• Benchmarked HiFi phasing tools (WhatsHap, HapDup, trio-binning) across 50 GIAB samples; results informed SMRT Link default recommendations.
• Automated T2T-style de novo assembly QC (hifiasm + Merqury + compleasm) for customer delivery pipeline.

Selected Publications

• Tremblay F. et al. “Real-time adaptive sampling for targeted long-read enrichment.” Nature Biotechnology, 2023.
• Tremblay F. et al. “Comprehensive SV benchmarking across HiFi and ONT R10 chemistry.” Genome Biology, 2022.

Code Quality Notes

Rust CLIs pass cargo clippy with zero warnings and are fuzz-tested with cargo-fuzz on BAM/FASTQ parsing paths. Go microservices include pprof endpoints and are load-tested with k6 before release. Snakemake pipelines containerised per-rule; AWS costs tracked per-sample via tagging and reported in automated Slack summaries after each run.