Larry H. Email & Phone Number

Belmont, Michigan, Us

La Jolla, Ca, Us

I developed the predictive modeling components of a web-based educational app intended for medical students and medical residents in radiation oncology.Responsibilities include:(1) Gather and process patient data for developing computer vision models based on deep learning(2) Train and evaluate convolutional neural networks (the deep learning models)(3) Develop software that utilizes model predictions to generate clinically meaningful outcomes(4) Mentor and delegate tasks to student assistants(5) Collaborate with other scientists and clinicians

Ann Arbor, Mi, Us

I was part of an interdisciplinary team in the Biomedical Clinical and Informatics Lab (BCIL) within the Michigan Center for Research in Critical Care (MCIRCC). In this proof-of-concept project, we developed the predictive modeling component of an early warning system intended for the intensive care unit (ICU) of the hospital. The system is intended to monitor cardiac patients and inform doctors if any patient is likely to develop serious complications during their post-surgical recovery in the hospital. Doctors could then apply interventional therapies to prevent any serious complications from occurring. I utilized patient data, signal processing techniques, and machine learning methods to develop the predictive modeling component of this early warning system.Achievements:(1) Developed 95% of the software for the data processing and modeling aspects of the project: data validation, feature generation and selection (i.e. dimension reduction), simple wrappers for predictive modeling algorithms.(2) Developed a strict version of data stratification that was most sensible for our data and predictive modeling goals.(3) Trained preliminary predictive machine learning models.(4) Improved one of the lab's signal processing methods for extracting statistical and morphological features from patient physiological waveform data.(5) Developed object-oriented code for processing physiological waveform data with wavelet packets: converted a published method to an implementation.(6) Co-authored a data science abstract for the Military Health Systems Research Symposium: classifying normal and abnormal heartbeats with a tensor-based machine learning method.(7) Documented progress in quarterly reports for the federal agency that funded our project.

Madison, Wi, Us

In short, I was developing new 3D MRI scans! It's a two-stage process: (1) develop software to get MRI scanners to collect data while people are inside of the scanner, and (2) develop more software to process and convert the raw MRI data into images (the raw MRI data do not look like anything that people would recognize as an image of the body). In MRI terminology, I developed MRI pulse sequence software and Fourier-based image reconstruction software.I maintained a few different scans, but the bulk of my efforts were spent on implementing and refining one specific scan that I will refer to as 'Rosette'. Clinical applications include detection of cartilage defects in the human knee, elbow, and hip joints. This same MRI scan may also help clinicians with breast cancer detection and diagnosis by improving visualization of breast tumors (think: see smaller structures with better clarity).I also conducted studies to compare the performance of our group's novel MRI scans against conventional, clinical MRI scans and presented findings at conferences. Throughout my time I trained several incoming students in the use of in-house MRI research tools, including operation of GE Healthcare MRI scanners.Achievements:(1) Implemented a software solution to a hardware memory deficiency on the MRI scanner. This memory deficiency had prevented installation of the new Rosette scan---the scanner just couldn't do it without the solution that I implemented.(2) Reduced primary cause of blurring in the MRI images: I utilized simulations to improve the MRI scanning method in a way that increased image-energy retention from 50% to 90%.(3) Developed a calibration method to eliminate electromagnetic physics-based image artifacts (i.e., “grey haze”, signal loss).(4) Equipped the Rosette scan with the capability to image the brain, breasts, and large joints of the human body (i.e., hips, knees, elbows).

London, England, Gb

As an Actuarial Analyst in the Retirement Practice, I worked on all aspects pertaining to valuation of defined benefit retirement plans, including special executive pension plans and retiree medical plans. This work included performing the annual pension data update process: cleaning and validating pension data, calculating pension liabilities using a parameter driven pension liability tool, analyzing the financial impact of various parameters, performing valuation integrity tests, and summarizing the results in actuarial reports and presentations for the company's clients.I also worked on: non-discrimination testing, pricing for union negotiations, annual benefit statements, ~4-5 pension benefit calculations per week, served as a point of contact for clients and responded to their inquiries, and prepared monthly billing statements to clients.Achievements:(1) Reduced annual data project time by 20% by automating manual steps of the data process.(2) Saved client 25% in cash contributions by resolving data issues.(3) Proved that office’s 3rd largest client complied with non-discrimination requirements.

B.A., Applied Mathematics & Physics

Phd, Medical Physics (Minor In Mathematics)