P a g e * *

Wolfgang Justice Black

Machine Learning Scientist Profile

adzjj9@r.postjobfree.com • 816-***-****

LinkedIn • San Francisco, CA

Summary

Machine Learning Scientist with 10+ years of experience in applying novel research to deliver innovative solutions to complex problems across multiple sectors. Hands on experience building and deploying machine learning models, deep learning algorithms, and statistical models to achieve business potential. Well-versed in Matlab, Python, TensorFlow, PyTorch and other libraries commonly used in Machine Learning. Demonstrated prowess in developing machine learning models, Computer Vision, Statistical analysis & data modeling, and Natural Language Processing. Technical Proficiencies

Data Science: Python, Matlab, Scala, SPARK, SQL, DataBricks/Jupyter, Unix Scripting, Cloud Computing, GitHub, AWS EC2, AWS S3, AWS ECR

Machine/Deep

Learning:

Computer Vision (Classification, Semantic Segmentation), NLP, CNNs, Supervised Learning, Unsupervised Learning, Semi-Supervised Learning, pipelines, Matplotlib, MlLib, Seaborn, Audio Classification, Docker, Research and Development

Education

PhD in Mechanical Engineering, Dissertation: Study of Magnetohydrodynamic Effects For The Richtmyer-Meshkov Instability University of Missouri

M.S in Mechanical Engineering, Dissertation: A preliminary study of shock driven multiphase hydrodynamic instabilities. University of Missouri

B.S in Mechanical Engineering

University of Missouri

Career Experience

Acorns, Irvine, Ca 03/2021 – Present

Sr. Data Scientist, Applied ML

Apply unsupervised learning techniques and integrate third-party data sources to create prototype model resembling product, increasing new user monthly subscriptions, and ensuring long-term sustainability. Implement DataBricks MLFlow and Model Registry to establish model pipelines and monitor performance drift, utilizing DataBricks to perform batch inference. Develop tooling for data science, analytics, and feature engineering following CI/CD best practices.

• Design a customer facing LLM API prototype for product use and support exploring frameworks around hallucinations like LangChain, Nvidia’s NeMo, and LLMs including ChatGPT3.5 and LlaMa

• Successfully created internal personalization framework utilizing unique multi-model architecture, facilitating prioritization of model outputs and customized treatment of new users with minimal data. The framework is highly adaptable and functions as an efficient communication engine during the initial 30-day user engagement period.

• Developed cutting-edge deep learning framework to forecast user-level time-series data, offering enhanced insights into new user lifecycle events both within and outside of Acorns system.

• Explore Deep Learning Solutions for Customer Support Audio data to reduce customer friction, decrease support costs, and generate internal auditing tool. Poised to reduce support costs by 25% and reduce wait times by 10%

• Created six models to effectively guide customer-centric communications, resulting in a notable 13% improvement in cohort survivability.

• Mentor Data Scientist and Junior Analyst in Machine Learning Development, Git Practices, and DataBricks Model Deployment

P a g e 2 5

JUUL Labs, Inc, San Francisco, CA 09/2018 – 12/2020 Senior R&D Scientist; Simulation Validation Manager Led and managed 4-person engineering R&D team which designed and executed multi-physics experiments, conducted CFD simulations using Flow3D, EXN/Aero, and SolidWorks Flow, and established effective protocols for prototype development. Management style 40/60 player-coach, with 40% of time spent between different CFD platforms and in the lab designing and vetting experiments and 60% of time opening cross-functional relationships and increasing team visibility and impact. Conducted extensive research on surface tension within high-fidelity Flow3D simulations, investigating impact of liquid constituents, gravity, and fluid-solid interactions, and validated findings through experiments for the Gen II product and Gen III prototype. Strengthened in-house CFD capabilities by creating validation datasets and analytical models, with emphasis on interplay between fluid dynamics and thermo-electric effects.

• Demonstrated expertise in simulation-based design by quickly exploring fluid-solid interactions and conjugate heat transfer using EXN/Aero and SolidWorks Flow for experimental prototypes, resulting in a 50% reduction in prototyping expenses and over a 50% increase in design efficiency.

• Utilized Computer Vision Fundamental techniques for analyzing IR Thermography data providing pixel to pixel comparison against First Principals CFD. Findings reduced prototyping time by 20% and increased simulation accuracy by around 5%

• Accomplished 4 patent filings, 2 complete conversions, and authored 3 conference proceedings and white papers within a span of 2 years through dedicated research efforts. Livermore National Laboratory, Livermore, CA 04/2017 – 05/2019 Graduate Researcher

Simulated multi-physics hydrocodes to generate datasets for machine learning of turbulence, reducing parameter sets for sub grid scale closure equations. Investigated Magnetohydrodynamics, multi-fluid, and multi-phase turbulence in simulations utilizing ARES, multi-physics hydrocode, resulting in 6 published research papers and conference proceedings. Worked cross- functionally with university groups and LLNL lab scientists to generate new avenues of research and to harden code capabilities. Generated bash shell scripts for automated experiment design and python packages for standardized simulation analysis.

Los Alamos National Laboratory, Los Alamos, NM 05/2015 – 04/2019 Graduate Researcher 0

Investigated Magnetohydrodynamics, multi-fluid, and multi-phase turbulence in simulations utilizing FLAG, multi-physics hydrocode, resulting in 8 published research papers and conference proceedings. Mentored several undergraduate, masters, and PhD students in Computational Physics and Research. Partnered with Laboratory Scientists to generate grants, new research supporting main Pillars of Research. Developed new physics packages to expand FLAG capabilities as well as python packages to support standardized analysis pipelines. Deep Learning/ML projects

LearnBot: Utilize LangChain LLM with ML specific knowledge to assist in research GitHub: [WIP]

Summary: finetune GPT3.5 with foundational articles covering Computer Vision, NLP, SEP, Transformers, and textbooks for PyTorch, Tensorflow, and Python to develop a Conversational Agent to assist in future research projects with an exposed API LLMDocExplorer

GitHub: https://github.com/wolfgangjblack/LLMDocExplorer Summary: Use LangChain framework to create a LLM model with context. Context is provided by a Chroma Vectorstore that is saved to disk. Currently ChatGPT 3.5 is the LLM supported Semi-Supervised Learning for Disease Classification on Optical Tomography Coherence Data GitHub: https://github.com/wolfgangjblack/semisup-learning-oct Summary: Designed and build a Semi-Supervised Learner based of SimCLR method to compare to traditional transfer learning for Optical Tomography Coherence Data and deployed containerized solution to AWS P a g e 3 5

ProtCNN/ProtENN System for Amino Acid Protein Sequence Data GitHub: https://github.com/wolfgangjblack/protein_seq_classification Summary: Designed and build a flexible containerized system to generate ProtCNN or Ensemble ProtCNN (ProtENN) model for protein family classification

Synthetic Data Generation and Classification with Google Text-to-Speech GitHub: https://github.com/wolfgangjblack/GttsAccentNet Summary: Generated a 3000 word-5 accent synthetic dataset and explored shallow CNN and transfer learning. Found that models faced challenges differentiated synthetic Canadian and US English accents but did provide a great starting model for real human accents

Speech Emotion Recognition utilizing Transfer Learning GitHub: https://github.com/wolfgangjblack/DeepEmo

Summary: Designed and build a flexible containerized system to generate ProtCNN or Ensemble ProtCNN (ProtENN) model for protein family classification

Next Best Action Meta-Model for New Users

Summary: Utilizing First Party and Third-Party Data designed N forecasting models fed into a recommender system to determine the Next Best Action a user can take to increase survival rates. Data combines demographic, Behavioral, and financial behaviors.

Sequence-to-Sequence prediction for modeling natural in-app behavior Summary: Design an ensemble system comprised of Recurrent Neural Networks and 1D CNNs to identify communication strategies and predict in-app/ecosystem behavior for user cohorts. System can identify rough window of screen exposure and in-app lengths.

Professional Training

Research and Development Data Scientist II, March 2021 – Present Acorns – Research and Development Senior Scientist and Simulations Validation Manager, September 2018 – December 2020 Juul Labs – Research and Development

Research Assistant and Laboratory Manager, August 2014 – September 2018 University of Missouri – Mechanical Engineering, Fluid Mixing Shock Tube Laboratory

Teaching Assistant, August 2016 – May 2018

University of Missouri – Mechanical Engineering, ThermoFluid laboratory Summer Scientist Intern, May 2018 – May 2019

Lawrence Livermore National Laboratory, Weapons Complex Integration, Design Physics Summer Scientist Intern, May 2017– August 2017

Lawrence Livermore National Laboratory, Weapons Complex Integration, Design Physics Summer Scientist Intern, May 2016 – August 2016

Los Alamos National Laboratory, X-Computational Physics Division, Lagrangian Codes Group Summer Scientist Intern, May 2015 – August 2015

Los Alamos National Laboratory, X-Computational Physics Division, Lagrangian Codes Group Teaching Assistant, August 2014 – May 2015

University of Missouri – Mechanical Engineering, Thermodynamics Undergraduate Research Apprenticeship Program Intern, May 2014 – August 2014 University of Missouri – Mechanical Engineering

Undergraduate Research Assistant, January 2013 – May 2014 University of Missouri – Mechanical Engineering, Oscillating Heat Pipe and Heat Transfer Laboratory P a g e 4 5

Licenses & Certifications

FourthBrain Machine Learning Cohort 7 Bootcamp Date of Completion: August 25, 2022 Capstone: Transfer Learning and Semi-Supervised for Retinal Optical Coherence tomography Structuring Machine Learning Projects Date of Completion: January 25, 2022, DeepLearning.AI Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization Date of Completion: January 21, 2022, DeepLearning.AI Neural Networks and Deep Learning Date of Completion: January 13, 2022 DeepLearning.AI Complete SQL Bootcamp 2021 Date of Completion: January 22, 2021 Jose Portilla, Udemy Machine Learning January 15, 2021, Stanford University Convolutional Neural Networks Expected Date of Completion: February 6, 2022 DeepLearning.AI Sequence Models Date of Completion: Feb 15, 2022, DeepLearning.AI Research Activities & Presentations

Morgan, Brandon E., and Wolfgang J. Black. "Parametric investigation of the transition to turbulence in Rayleigh–Taylor mixing." Physica D: Nonlinear Phenomena 402 (2020): 132223. Black, Wolfgang J., Roy C. Allen, W. Curtis Maxon, Nicholas Denissen, and Jacob A. McFarland. "Magnetohydrodynamic effects in a shock-accelerated gas cylinder." Physical Review Fluids 4, no. 4 (2019): 043901. Morgan, Brandon E., Britton J. Olson, Wolfgang J. Black, and Jacob A. McFarland. "Large-eddy simulation and Reynolds- averaged Navier-Stokes modeling of a reacting Rayleigh-Taylor mixing layer in a spherical geometry." Physical Review E 98, no. 3

(2018): 033111.

Allen, Roy, Wolfgang J. Black, and Jacob A. McFarland. " Development and diagnosis of an atmospheric pressure plasma torch for investigating magnetohydrodynamic instabilities." JPhysD-118370.R2. Middlebrooks, John B., Constantine G. Avgoustopoulos, Wolfgang J. Black, Roy C. Allen, and Jacob A. McFarland. "Droplet and multiphase effects in a shock-driven hydrodynamic instability with reshock." Experiments in Fluids 59, no. 6 (2018): 98. Black, Wolfgang J., Nicholas A. Denissen, and Jacob A. McFarland. "Particle Force Model Effects in a Shock-Driven Multiphase Instability." Journal of Shockwaves. DOI: 10.1007/s00193-017-0790-0 Black, Wolfgang J., Nicholas A. Denissen, and Jacob A. McFarland. "Evaporation Effects in Shock-Driven Multiphase Instabilities." Journal of Fluids Engineering 139.7 (2017): 071204. McFarland JA, Black W, Dahal J, Morgan B. 2015. “Computational Study of the Shock Driven Instability of a Multiphase Particle- Gas System”. Physics of Fluids. 28 (024105)

Zhang F, Wilson M, Black W, Winholtz RA, Ma H.B. “Effects of Hydrophilic Nanostructures Cupric Oxide (CuO) Surfaces on the Heat Transport Capability of a Flat Plate Oscillating Heat Pipe”. Journal of Heat Transfer. (HT-14-1678) McFarland JA, Reilly D, Black W, Greenough JA, Ranjan D. 2015. “Modal Interactions between a large-wavelength inclined interface and small-wavelength multimode perturbations in a Richtmyer-Meshkov instability”. Phys Rev E. 92 (013023) Presentations

Black, W., Alston, W., Holloway, G., Zhang J.T. October. 2019. “Computational simulation of aerosol generation and temperature regulation performance of nicotine salt pod system”. CORESTA Meeting, Smoke Science/Product Technology 2019 Hamburg, Germany.

Black, W.J. Aug. 2018. “Parametric investigation of transition to turbulence in Rayleigh Taylor mixing”. Lawrence Livermore National Laboratory – Internal Presentations. Livermore, CA. J Middlebrooks, CG Avgoustopoulos, WJ Black, RC Allen, JA McFarland. 2018 “Experimental Technique for the measurement of velocity and droplet lag distance in a shock accelerated multiphase system”. Bulletin of the American Physical Society 63. P a g e 5 5

Morgan, B., Black, W.J., and J.A. McFarland. Nov. 2017. “Large-Eddy Simulations of a Rayleigh-Taylor Instability in a Convergent Geometry”. 70th American Physical Society – Division of Fluid Dynamics Meeting. Denver, CO. Black, W.J., Denissen, N.A., and J.A. McFarland. Nov. 2016. Probabilistic Events in the Shock Driven Multiphase Hydrodynamic Instability 69th American Physical Society – Division of Fluid Dynamics Meeting. Portland, OR. Middlebrooks, J., Black, W.J., Avgoustopoulos, C., Allen, R., Kathakappa, R., and J.A. McFarland. Nov. 2016. “Design and Construction of a Shock Tube Experiment for Multiphase Instability Experiments”. 69th American Physical Society – Division of Fluid Dynamics Meeting. Portland, OR.

McFarland, J.A., Black, W.J., Dahal, J., and Morgan B. Nov. 2015. “Shock Driven instability of a Multi-phase Particle-Gas System.” 68th American Physical Society – Division of Fluid Dynamics Meeting. Boston, MA. Black, W.J., Denissen, N.A., and J.A. McFarland. Nov. 2015. “Shock Wave Interactions in Multi-Phase Particle Systems Characterized by Various Interfaces”. 68th American Physical Society – Division of Fluid Dynamics Meeting. Boston, MA.

Contact this candidate