Tim Dunn Email & Phone Number

Ann Arbor, Michigan, United States

Ann Arbor, Michigan, United States

My research focused on evaluating germline variant calling for whole-genome sequencing (WGS). I released vcfdist, the first tool to perform accurate phased whole-genome evaluation of all variant types, from single-nucleotide polymorphisms (SNPs) to structural variants (SVs) up to 10,000 bases in size. [Nature Communications, 2023] [Genome Biology, 2024]Prior to that project, I invented an extension to the Smith-Waterman-Gotoh alignment algorithm to better align simple tandem repeat (STR) variants and released an open source tool called nPoRe. This work, in combination with re-training downstream machine learning (ML) variant callers, improved recall of insertion and deletion variants from 64% to 73%. [BMC Bioinformatics, 2023]My first project during graduate school was motivated by the COVID-19 pandemic. I designed an algorithm and hardware-based filter named SquiggleFilter for excluding non-viral reads in real time as a nanopore-based device (MinION) sequences a saliva sample. Our simulations estimate that this filter reduces sequencing cost by 38% because over 99% of DNA in a saliva sample is non-viral, and our filter manages to avoid much wasted sequencing and computation. [MICRO, 2021]𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: C++, Python, BASH, Cython, Git, Docker, samtools, JupyterNB, SystemVerilog

I fine-tuned open source artificial intelligence (AI) large language models (LLMs) using Python on patient portal messages to generate suggested responses for health care providers, and wrote a report detailing my efforts. To assist with further research, I generated a training dataset of 1.3 million patient-to-provider conversations.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: Python, Google Cloud, pytorch, LaTeX

Redmond, Washington, United States

I developed an Azure cloud pipeline for integrating long-read whole-genome sequencing (WGS) data in variant call format (VCF) with clinical health record data in FHIR format. This enables scalable and secure analyses which combine patient genetic and health information, leading to better understanding of disease treatment.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: Python, Git, Microsoft Azure, Azure Confidential Compute, HL7 FHIR, WDL, JupyterNB

Rome, Ny

I collaborated with two other interns to integrate a virtual Trusted Platform Module (vTPM) into OpenXT, a secure Xen-based hypervisor.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: Linux, BASH, BitBake, C

Potsdam, Ny

I worked with two peers to teach an introductory MATLAB programming class to 50 freshman Honors students. My senior year, I designed and taught a new curriculum based on student feedback from previous years.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: MATLAB, LaTeX

Potsdam, Ny

I worked in the TARS research lab with professors Sean and Natasha Banerjee, performing research in: remotely detecting finger swipe pressure using thermal camera (my Honors Thesis), mining bug report repositories, accelerating 3D-model alignment to a 2D image, and in optimizing MATLAB PointCloud generation.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: Python, C++, CUDA, MATLAB, JupyterNB, Git, Docker, LaTeX

Darmstadt, Germany

I performed research abroad through the DAAD RISE academic exchange service and wrote an open-source parallel solver for semidefinite optimization problems in C++.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: C++, CUDA, MATLAB, CMake

Los Alamos, New Mexico

I ported a computational x-ray imaging algorithm from CUDA C++ to OpenMP and implemented a conjugate gradient descent position refinement algorithm to improve reconstructed image quality.𝗧𝗲𝗰𝗵𝗻𝗼𝗹𝗼𝗴𝗶𝗲𝘀: C++, CUDA, CMake, OpenMP, Slurm

Doctor Of Philosophy - Phd, Computer Science, 4.00

Activities and Societies: Climbing Team member, Ballroom Dance Team member, Nordic Skiing club member, Backpacking club member, Triathlon.

Honors Bachelor Of Science (Bs), Electrical Engineering And Computer Science, 3.99

Activities and Societies: Captain of Varsity Nordic Skiing Team, President of Juggling Club, Webmaster for Honors Program, Crew Team.