PhD in Data Science/Physics, experience with Python, Machine Learning

Andrew T. Cadotte, Ph.D.

Ann Arbor, MI ***** ********@*****.*** 313-***-**** LinkedIn: https://tinyurl.com/4rdxsdb3 GitHub: github.com/cadottea

Summary

Experienced machine learning engineer with expertise in Python, C++, and TensorFlow. Skilled in developing large-scale simulations, perception systems, and data workflows. Developed and deployed solutions to high-complexity engineering problems, driving innovation in data analysis workflows. Self-supervised learner, experienced in model optimization, and transfer learning. Actively seeking an opportunity to apply technical expertise and problem-solving skills to cutting-edge projects in machine learning. Technical Skills

• Programming: Python, C++, SQL, Mathematica, Swift, Ruby

• AI/ML Tools: TensorFlow, Large Language Models, CoreML, YOLO, Huggingface, Pytorch. VectorDBs

• Simulation Tools: HOOMD-blue (GPU-accelerated Molecular Dynamics)

• Data Visualization: Tableau, Power BI

• Version Control: Git/GitHub

• Applications: Microsoft Office Suite, Slack, SharePoint Education

University of Michigan, Ann Arbor

Ph.D. Applied Physics, 2022

● Thesis: https://deepblue.lib.umich.edu/handle/2027.42/174308 University of Michigan, Ann Arbor

Bachelor of Science in Physics

Research Experience

Computational Self-Assembly with Prof. Sharon Glotzer, Ph.D., NAS Ph.D. Graduate Student, University of Michigan (50 hours/week)

● Pioneered the first reported self-assembly of a binary co-crystal of hard tetrahedra and octahedra, utilizing Python and HOOMD-blue to create and study the self-assembly phase diagram

● Used HOOMD-blue to conduct molecular dynamics simulations with size limits approaching supercomputing limit of >5 million particles to study the nucleation and growth of complex crystals

● Wrote Python code to self-assemble Penrose tilings with unprecedented quality by incorporating matching-rule equivalent enthalpic patches on hard particles

● Authored novel C++ order-parameter analysis for quasicrystalline systems, providing a new framework for structural classification. Applied YOLO for object detection on datasets >10M images Atomic Molecular Optics with Prof. Georg Raithel, Ph.D. Ph.D. Graduate Student, University of Michigan (50 hours/week)

● Collaborated extensively across institutional lines to build precision parts in step-by-step incremental improvements. Direct communication with dozens of companies to diagnose and resolve problems with previous design for the assembly and testing of the Bose-Einstein Condensate (BEC) II experiment

● Conducted MATLAB simulations to optimize apparatus design for high-resolution imaging, improving data fidelity for precision experiments

Atomic Molecular Optics with Assistant Prof. Aaron Leanhardt, Ph.D. Research Scientist, University of Michigan (20 hours/week)

● Designed and machined parts for a novel ytterbium BEC aviation gyroscope. Developed algorithms in C++ to control ytterbium ablation rates and model ytterbium molecular flow rates to calibrate BEC formation

● Designed and implemented a novel vibration detection instrument using a He-Ne laser and a glycerin-filled cavity to study the efficacy of floating optic research tables, providing critical quality assurance for the team’s design High Energy Particle Physics with Prof. Dante Amidei, Ph.D. Undergrad Research Scientist, University of Michigan (10 hours/week)

● Developed Python decision trees to analyze massive particle-collision data to parse out rare events. Part of the initiative to discover a new particle - the Higgs boson AI Projects/Open Source Work

Experience developing, fine tuning, and evaluating Large Language Models (LLMs) using Huggingface, PyTorch, and TensorFlow, with working knowledge of distributed training and model compression techniques Morel Hunter - Developed a YOLO-based deep learning model to detect morel mushrooms for application integration Whisper Optimization – Implemented low-latency, parallelized inference pipeline for Whisper (a 1B+ parameter transformer model) using whisper.cpp, targeting Apple’s M1 GPU. Gained hands-on experience with transformer inference bottlenecks, model quantization, and deployment constraints relevant to scaling LLMs. Audio Transcription Project - Python wrapper for the Whisper model using whisper.cpp, optimized for Apple’s M1 GPU Edgeword Finder - Developed a tool to analyze words based on edge pattern matching, useful for linguistic analysis Fourier Analysis of Sleep Data - Performed Fourier analysis on FitBit sleep data to uncover sleep cycle trends VPN Throttling SpeedCheck - Collects upload/download speed data with/without an active VPN and provides analysis Team Leadership & Project Management

● Managed teams of 3 graduate students on large-scale simulation projects, ensuring timely completion and peer-reviewed publication

● Facilitated cross-disciplinary collaboration between engineers and researchers to solve computational challenges

● Led a 4-month dissertation writing group, coaching 5 Ph.D. students to improve clarity and productivity Publication & Patent

● A.T. Cadotte et al. (In Prep). Layering: A novel real-space method for constructing quasicrystal tilings

● A.T. Cadotte et al. (2016). Self-assembly of a space-tessellating structure in the binary system of hard tetrahedra and octahedra. Soft Matter, 12

● I. Han, A.T. Cadotte et al. (2021). Formation of a single quasicrystal upon collision of multiple grains. Nature Communications, 12

● Miniature mechanical shutter: U.S. Patent No.: 9831754 Teaching Experience

Graduate Instructor University of Michigan

● Taught Python-based data analysis and Excel for introductory physics labs, instructing 50–100 students annually.

● Achieved certifications in Laser Safety, Radiation Safety, and Machine Shop Safety MallyACT ACT/SAT Math, Science, English, and Reading Tutoring (2023-present)

● 40-60 hours/week

Society of Physics Students University of Michigan

● 1-2 hour sessions, 50 students a year - Varsity Tutors, Preply, Wyzant online tutoring Applied research LLM self-supervised learning large language models transformer models quantization gradient checkpointing fine-tuning pretraining model compression distributed training 10B+ parameters transformer inference bottlenecks Lightning AWS, Ultraclusters, RLHF instruction tuning token embeddings vector databases ICLR NeurIPS ICML NLP gitflow Apache Airflow, dbt, Glue, Dataflow, BigQuery, Snowflake

Contact this candidate