Tim Matchen Email & Phone Number

With Mynd's merger with Roofstock, I'm excited to be continuing my role as the data scientist for our Generative AI team.

I'm working on the Generative AI team at Mynd to revolutionize service in property management.

Provided analysis, maintained models, and designed/implemented testing strategies across a diverse set of projects in property management.

Identified novel applications for LyteLoop's core technology and researched, pitched, and developed business and development plans for new project.Established project development requirements and managed team to meet development milestones. Implemented a hybrid project management plan integrating Waterfall and Agile methodologies. Led collaborations with multiple external partners.Spearheaded writing and submission of grant proposals and RFI responses across several projects within LyteLoop.

Great Neck, New York, United States

Identified internal opportunities to supplement research efforts with machine learning and artificial intelligence and provided expertise on machine learning topics.Modeled free space intrasatellite beam propagation and implemented a Bayesian optimization scheme to identify the optimal beam mode for transmission under LyteLoop's design requirements.Processed, evaluated, and visualized lab results for multiple spatial division multiplexing experiments.

Santa Barbara, California, United States

Identified disparity in freshman student performance based on background and researched, developed, and taught new course to help address said disparities.Developed, pitched, and taught an honors section for the introduction to MATLAB class with a specific emphasis on machine learning.Created a project-based curriculum culminating in a midterm project in which students built an artificial neural network from scratch and a final independent project in which students developed working algorithms using an advanced topic in machine learning of their choice, including projects related to Deep Q-Learning and Monte Carlo Tree Search.

Santa Barbara, California, United States

As part of UC Santa Barbara's Research Mentorship Program (RMP), mentored six high-achieving high school students over three summers. Set milestones for students to track progress and developed projects of an appropriate skill level. Provided written and verbal feedback on final paper, oral presentation, and poster.Drafted a grant proposal to receive university funding to support undergraduate researchers in the Moehlis lab group. Grant was subsequently approved by the university, allowing undergraduates to be compensated for their research time.Mentored four undergraduate students on separate occasions. Coordinated weekly meetings, provided research direction to students, and worked with the students to develop realistic timelines for their research. Assisted one student with preparing a paper for journal submission.

Investigated high-efficiency control strategies to achieve neuronal desynchronization and clustering. I run numerical simulations using Matlab, Python, and C++ as well as working on the underlying theory behind potential control methods. Worked with Professor Jeff Moehlis.Developed adaptive-closed loop algorithms for modifying the firing patterns of populations of coupled neurons from a single common input stimulus. Expanded the work to open-loop, stochastic neuron networks to achieve arbitrary neuron clusters. Trained a neural network to discriminate between pathological and benign behavior using collected clinical data from Parkinson's patients. Implemented an autoencoder for feature extraction, then used the encoded features for prediction. Prediction error showed as much as a 35\% improvement over baseline null hypothesis against validation set.Developed an innovative, neural-network-based approach to system identification applicable to a wide class of dynamical systems without requiring assumptions about solution characteristics. The algorithm successfully identifies underlying governing equations for a wide array of physical and biological example problems, even in the presence of noise, and provides highly-parsimonious analytical expressions for complex dynamics, including chaotic systems.Contributed to additional research on other topic areas, most notably system identification via Koopman Operator Theory and underactuated control theory.

Santa Barbara, California, United States

Led sections, graded assignments, and held office hours and review sessions for undergraduate courses, primarily introduction to programming in MATLAB.Introduced an in-class "live blog" during lecture to provide supplementary insight to course lectures.Helped coordinate and provided feedback to undergraduate mentors who assisted students in the course.

Oversaw operations for the country's 3rd oldest humor magazine, planning issue themes and developing publicity strategy as well as contributing articles and serving as an editor.

Oversaw the restructuring of the magazine's organizational design and began learning the necessary skills to take over as Chairman.

Served as the chief organizer for the 1st Annual National Intercollegiate Humor Conference and continued to contribute for the 2nd and 3rd humor conferences in succeeding years.

Conducted independent research on using camera motion to disambiguate object trajectories from captured video. Built Monte Carlo simulations to validate results in Matlab and collected experimental data to extend research.Won "Outstanding Poster Presentation" at the 2013 Los Alamos Student Symposium. Criteria for the award were excellence in both poster design and clarity as well as a two-minute oral presentation of the poster.Wrote, submitted, and subsequently received acceptance of a conference paper on estimating trajectories from captured video footage. Presented paper at a session of the 2014 IEEE Aerospace Conference.

3D3A Lab

Automated the collection of data in speaker directionality measurements for 3D audio research. Automation required interfacing a high-torque turntable with a standard Apple computer. Through experimentation, identified the appropriate input commands to set the turntable angular position to any prescribed angle for directionality measurements.Integrated automated testing routine into a larger testing interface built in Mathematica. Once integrated, testing automation reduced the time required to complete a full set of measurements for a speaker from over 30 minutes to under 5 minutes.

Doctor Of Philosophy (Ph.D.), Mechanical Engineering

Activities and Societies: Certificate in Neuroengineering

Bse, Mechanical And Aerospace Engineering

Activities and Societies: Program in Engineering Physics Princeton Charter Club Games Chair