Melanie Kirsche

• Applying state-of-the-art machine learning techniques to improve genetic variant calling from high-error genome sequencing datasets
• Developing NextFlow pipelines for Google Cloud to measure and report variant calling accuracy across a range of genomics contexts and leading cross-team experiments to evaluate the impact of sequencing technology changes on bioinformatics findings
• Developed a graph-based algorithm to cluster and phase assembled genomic sequences into haplotypes

• Designed and implemented state-of-the-art bioinformatics pipelines using NextFlow, Python, and R for large-scale analysis through AWS Batch
• Analyzed whole-genome and whole-exome sequencing from diverse human populations, including variant calling, imputation, genotyping, quality control, and visualization
• Integrated pipeline results with MySQL databases and helped develop database schemas

    Research areas:

• Designed algorithms and data structures for population-scale genomic structural variant data, and released software around them
• Contributed to the first complete human genome assembly and assessed its impact on variant calling
• Designed data structures to accelerate genomic read alignment

• Implemented Louvain clustering algorithm and integrated it into the CellRanger single-cell RNA pipeline
• Implemented a Rust API for the STAR RNA read aligner and used it to improve CellRanger’s efficiency

• Implemented algorithm for landmark recognition in Bing Images

• Automated quantifying and categorizing of bandwidth usage in the Google Search Android application

• Designed and implemented an object tracking computer vision algorithm to analyze cell state trajectories in longitudinal microscopy data

• Implemented an iOS lab device remote control proxy client/server for running automated application tests