George M. Mastorakos, MD, MS
- georgemasto@gmail.com
- linktr.ee/georgemasto
- Scottsdale, AZ
- ⚕️ Medical graduate w/ experience in end-to-end clinical AI software product development (ideation to post-market surveillance), real-time ML clinical validation workflows, cloud computing & data management, and clinical informatics
- 👨💻 Passionate about meshing biomedical background with AI development experience to bridge gaps between clinical stakeholders and AI engineering endeavors, creating clinical software products that enhance the life of patients at home and clinicans on the front-lines
- 🗣 Excited to build, validate, and deploy explainable AI systems to clinicians, researchers, patients to further demystify the "black box" of AI systems
- 🌱 Special interests in biomimicry (the emulation of nature's genius), quantum computing, and pharmacognosy
Experience
Senior Clinical AI Solution Architect
Clinical AI Software Development
Clinical Software Developer
- Design, code, train, review, and deploy convolutional neural networks for automated detection of brain tumors and other neurological lesions
- Design and build customizable AI model performance assessment tools for product reviews with leadership and clinical stakeholders
- Lead clinical AI product validation sessions with several radiologists
- Document AI product development and V&V processes per US (i.e. FDA) and international (e.g. CE) regulatory requirements
- Curate, assess quality of, and preprocess thousands of medical images (e.g. MRIs) across 3 product lines
- Maintain and automate company-wide cloud data storage, versioning, and traceability
- Produce, review, and test high-quality, production-level, easy-to-maintain Python code
- Coordinate ~1 brainstorm per week, providing clinical and engineering input for feature scoping and prototyping
- Write ~1 clinical blog post per month for company website
Domi: Machine Learning for Patient Message Triage
Team Lead/Programmer
- Mission: Automate the laborious processes of triaging and routing incoming patient portal messages
- Collaborated with a team of computer engineers to create a multi-model ensemble classifier (0.97 AUC) and named-entity recognizer
- Recieved competitive CoDE Grant ($50,000) from Mayo Clinic
- Led an annotation team of 15 students to generate a training corpus of 10,000+ patient portal messages
- Epic EHR and Google Cloud Platform integration ongoing
COVID-19 Triage and Virtual Survey
Clinical Lead/Programmer
- Mission: Train & apply ML models to quickly and effectively triage patient COVID concerns
- Trained multimodal models on corpus of 2 million patient-generated messages related to COVID
- Designed COVID-specific labeling scheme (e.g. swab test results, cancellations, pandemic anxiety) to an incoming stream of patient messages
- Deployed at height of pandemic with internal hospital disaster response team
Hand Digit Joint Spacer Product Development
Clinical Advisor
- Mission: Develop a novel hand digit joint spacer that eliminates the need for amputation in patients with bone infections
- Leveraged biomimicry knowledge to produce novel joint spacer designs
- Reviewed designs, engineering prototypes, and integrated clinical feedback into development of product with a collaborative group of 7 engineering & business students at ASU
- Navigated patenting and venture processes, including provisional patent establishment to building patentable claims
Molecular & Cellular Immunoproteomics Research
Research Fellow
- Mission: Investigate NK cell receptor family KIR structure, localization, and binding dynamics (Mentors: Eric O. Long; Sumati Rajagopalan)
- Designed and conducted wet-lab experiments in which random mutagenesis of KIR proteins was applied after which transfection, localization, and ligand binding assays were performed
- Operated confocal laser-scanning microscope to detect bioluminescent antibody protein interactions with specific mutants of interest
- Programmed SCWRL4 scripts to model mutant proteins of interest and side chain dynamics
- Built Python pipeline for reproducible modeling of protein—ligand interactions
Structural Characterization of Novel Cancer Targets
Research Fellow
- Mission: Evaluate drug-screen inhibitors of novel cancer target ITPA via crystallization and 3D structural analysis (Mentors: Pål Stenmark; Ronnie Berntsson)
- Engaged in wet-lab preparation of crystallography experiments, from bacterial expression/purification of protein to carrying out crystallization trials
- Obtained crystals, sent to synchrotron, collected diffraction pattern data to 2.0Å resolution
- Postprocessed and refined data using CCP4 and Refmac5 software tools; built models and fit to previous ones using Coot
Be the Match® on Campus
Co-Founder, President
- Mission: Advertise need of bone marrow donors, raise funds for transplant costs, and sign up potential donors
- Collaborated with team of 10 students to educate community members about blood cancers and the necessity of bone marrow transplants
- Created automatic registration platform, registering > 3000 students as potential marrow donors
- Led a cross-country road trip to educate other institutions about Be the Match's mission
Selected Publications
➔ 📜 Google Scholar for more
Journal Articles
Huang M, Mastorakos GM, Khurana, A, et al. (2022) “Identification of Transportation Barriers in Patient Portal Messages via Deep Semantic Embeddings and Clustering”. Stud Health Technol Inform. DOI: 10.3233/SHTI220188.
Huang M, Mastorakos GM, Khurana, A, et al. (2022) “Patient Portal Messaging for Asynchronous Virtual Care During the COVID-19 Pandemic: Retrospective Analysis”. JMIR Hum Factors. DOI: 10.2196/35187.
Mastorakos GM, Khurana A, et al. (2021) “Probing Patient Messages Enhanced by Natural Language Processing: A Top-Down Message Corpus Analysis”. Health Data Science. DOI: 10.34133/2021/1504854.
Tafti AP, Fu S, Khurana A, Mastorakos GM, et al. (2019) “Artificial intelligence to organize patient portal messages: a journey from an ensemble deep learning text classification to rule-based named entity recognition”. 2019 IEEE Bioinformatics and Biomedicine (BIBM), 1380–1387. DOI: 10.1109/BIBM47256.2019.8982942.
Patel P, Ivanov D, Bhatt S, Mastorakos GM, et al. (2019) “Low-Cost Virtual Reality Headsets Reduce Perceived Pain in Healthy Adults: A Multicenter Randomized Crossover Trial”. Games for Health Journal. DOI: 10.1089/g4h.2019.0052.
Pines AR, Khurana A, Mastorakos GM, et al. (2018) “A Prognosis for Health Systems Science Courses: Observations From Current Students”. Academic Medicine. 93(10), 1434–1436. DOI: 10.1097/ACM.0000000000002369.
Pfaller CK, Mastorakos GM, Matchett WE, et al. (2015) “Measles Virus Defective Interfering RNAs Are Generated Frequently and Early in the Absence of C Protein and Can Be Destabilized by Adenosine Deaminase Acting on RNA-1-Like Hypermutations”. Journal of Virology. 89(15), 7735–7747. DOI: 10.1128/JVI.01017-15.
Conference Presentations
Mastorakos GM. (2022) "Segmentation of pre- and post-treatment glioma tissue types including resection cavities". Society of Neuro-Oncology 2022. Tampa, FL.
Mastorakos GM. (2021) "Patient Asynchronous Response to Coronavirus Disease 2019 (COVID-19): A Retrospective Analysis of Patient Portal Messages". AMIA 2021. San Diego, CA.
Mastorakos GM. (2021) "Analysis of Patient Portal Usage among Patients and Providers from 2010-2017: A Single Institutional Cohort Study". AMIA 2021. San Diego, CA.
Mastorakos GM, Khurana A, Bolen E, et al. (2019) “Developing a fully annotated textual dataset to advance autonomous patient message summarization and classification”. Artificial Intelligence Symposium. Mayo Clinic, Phoenix, AZ.
Mastorakos GM and Khurana A. (2019) “Domi: Automated Patient Message Organization and Virtual Triage System”. Academic Excellence Day. Mayo Clinic, Phoenix, AZ.