Matthew Woods

MATTHEW WOODS, PH.D. San Jose CA • (617) 510 – 6399 •- Professional Summary Machine Learning Engineer | Research Scientist | Technology Leader, Innovator & Strategist Promoting technological transformation that delivers powerful, revenue generating solutions. ! Solutions-driven Technology Leader with expertise designing, developing, and driving Machine Learning and Artificial Intelligence initiatives through strategic assessment or large-volume data, encouraging innovation, maintaining focus on delivery, and motivating high-performing teams. Identifying innovative approaches to business development and communicating initiatives with executives, boards, and stakeholders to remain ahead of technology trends. ! Successful track record of creating results by transforming technology to position growing organizations and startups. Designing outcome-focused technical strategies and engineering processes to maximize opportunities. Skilled at connecting the dots between technology, business and profitability while providing seamless execution with a focus on mentorship, integrity and service. Skills • Innovative Machine Learning Strategies • Providing Direction, Vision & Leadership • Software Development and Signal Processing • Efficient Technology Implementation • Patent Writing and R&D process • Prioritizing Multiple Projects • Aligning Technology to Business Goals • Commitment to Customer Service • Developing & Supporting Technology Teams • Thinking & Planning Strategically • Collaborating with Key Stakeholders • Maximizing Profitability & Reducing Costs Work History Consultant / Freelance Machine Learning Research Engineer, 6/2021 to Current Independent – San Jose, CA • Designed and built Machine Learning production systems from the ground up for various clients. • Delivered highly performant Machine Learning systems for clients in a range of industry verticals including medical diagnostics and automotive. • Guided data collection and annotation teams. • Mentored Machine Learning Engineering teams. • Built Machine Learning and Computer Vision models and systems utilizing image, language, spectrographic, tabular, and other data types. Principal Software Development Engineer, Artificial Intelligence, 01/2021 to 06/2021 Traptic – Sunnyvale, CA • Hired to utilize Artificial Intelligence knowledge and strong technical skills to help grow Traptic, which is an early-stage startup with around 20 employees working on a strawberry picking robot. • Led the development of the robot's visual system software using Deep Learning and Computer Vision, including coding in Python and Bash for training, tuning, and deploying to production customized Convolutional Neural Networks based on Resnet and Retinanet architectures. • Contributed the architectural modifications and custom loss functions for the neural networks that allowed the inference of multiple targets from images with semi-supervised training. • Executed the necessary modifications to the inference engine allowing it to serve inferences for the newly developed targets. • Enabled the training on multiple GPU-endowed machines by containerizing the training routines with Docker. • Championed the delivery of a mathematical model of harvest yield that allowed tuning aspects of the visual and robotics systems to maximize the robot's strawberry picking rate. Founder / Machine Learning Scientist, 10/2019 to 1/2021 Coniveo – San Jose, CA • Set strategy and direction to further develop, improve, and commercialize the driver state monitoring system that was initially designed and developed at SAIC-IC. • Leveraged robust knowledge and skills in combining signal processing and machine learning to successfully produce a technology that monitors a vehicle operator's drowsiness level through real-time vital sign monitoring, driver-facing camera, and driver interaction with vehicle controls. • Engineered this technology by designing software to automate model inference, transforming data science prototypes, and applying appropriate algorithms and tools. • Performed various responsibilities to achieve objectives such as running tests, performing statistical analysis, interpreting test results, and documenting machine learning processes. • Successfully managed networking activities as well as pushing funding opportunities and exploring and connecting with the Bay-Area startup ecosystem. Senior Machine Learning Engineer, 01/2018 to 09/2019 SAIC Innovation Center – San Jose, CA • Served as the Senior Machine Learning Engineer at the The Innovation Center at SAIC (The Shanghai Automotive Industry Company) which is a research center for technological innovation in vehicle-embedded software. • Worked in the Human Machine Interface group and collaborated with cross-functional technical teams to plan, design, and develop innovative technologies utilized in vehicle interiors. • Received the Innovation Center's Outstanding Service Award in 2018. • Filed SAIC-IC's first patent application. • Provided direction, vision, and leadership to the Artificial Intelligence team within the HMI group, including supervising 2 Machine Learning Engineers and 1 Data Engineer, training, coaching, and assessing technical performance, accomplishing team targets, and producing positive results. • Oversaw projects and served as an individual contributor delivering novel core technologies, algorithms, and software to Drowsiness Detection System, Smart HVAC, and Driver Destination Prediction projects. • Ideated, strategized, planned, and executed the Driver Drowsiness Detection project focusing on producing a working prototype of an in-vehicle drowsiness detection system. o Employed machine learning and computer vision to detect drowsiness of the driver from a combination of sensor sources including a driver-facing camera, driver biometric measurements (in-vehicle vital sign monitoring), and steering wheel movements. o Designed and conducted experimental setup to gather training data for the drowsiness detection system; observed 50 subjects in a two-hour session using a driving simulator and recorded the data from the driver's interaction with the driving simulator, from a driver-facing camera, and from biometric hardware to capture EEG, ECG, and driver respiration. o Combined 3 data sources including signal processing (FFT, wavelets, etc.), computer vision (extraction of facial features from the video feed), and machine learning components. • Managed the development of the Smart HVAC system which is a technological product that anticipates the driver's preferred environmental settings and executes preemptive adjustments to the HVAC Unit. The machine learning core learns to anticipate an individual's preferences on the basis of in-vehicle sensor measurements, such as temperature, and sunlight, and HVAC unit settings, such as current airflow. • Solely handled the mapless destination prediction project focused on the development of an algorithm that predicts the three most likely destinations of the driver and updates these forecasts as the trip progresses. The algorithm combines gradient boosted tree regression, kernel density estimation techniques, and statistical modeling to predict destinations based on the current time and location of the vehicle. Senior Data Scientist, 07/2015 to 12/2017 Venafi – Palo Alto, CA • Worked for a cyber-security company committed to protecting clients from attacks that exploit vulnerabilities in the keys and certificates that are used to establish identity during online communication. • Played an integral role in the timely accomplishment of several R&D projects and the approval and filing of four patent applications. o Managed the patent application of machine-learning (supervised learning) methods for the evaluation of the reliability of TLS certificates. Explored various methods including Elastic Net Regularized Logistic Regression, Deep learning Neural Network architectures, and Gradient Boosted Trees. o Filed a patent application for work developing an anomaly detection method (using unsupervised learning and density estimation) applicable to keys and certificates. • Utilized strong analytical skills to evaluate the comprehensive data collected at Venafi, identifying current trends and regularities in the security landscape, and used a large collection of robust non-linear regression models to produce an indicator to guide the aspects of Venafi's sales strategy. • Detected new problem domains with the potential to add value, designed statistical and Machine Learning methods to solve these problems, and implemented methods in Python, Spark, and R. Leveraged high-volume datasets in the hundreds of gigabytes range, employed high-performance cluster computing on Amazon cloud resources, and used a range of big-data tools and methods. Machine Learning and Bioinformatics Consultant, 12/2014 to 05/2015 Transplant Genomics – Woodlands Hills, CA • Assessed, designed, and developed machine learning algorithms and feature selection approaches to produce a blood test that can determine the likelihood of kidney transplant rejection earlier than existing technologies by measuring the expression of genes in patients' blood. • Delivered and performed R code to detect and classify the molecular signature and predict the transplant rejection likelihood while producing graphical displays of the results and methods • Applied Machine Learning methods such as Naïve Bayes classifiers and Logistic Regression, and feature selection methods such as univariate methods (t-tests, F-tests) and multivariate method Maximum Relevance Minimum Redundancy (MRMR). Principal Bioinformatics Scientist, 10/2012 to 03/2014 SynapDx – Lexington, MA • Worked on the R&D of machine learning systems to classify patients as positive or negative for Autism Spectrum Disorder, software development to build a robust data processing pipeline, algorithm development for normalizing and de-noising gene expression data, and A/B testing and analytics targeting the improvement of the laboratory protocols. • Used different machine learning techniques to identify the signature of autism in blood gene expression which involved extensive investigations into the feature selection aspect of the problem. Compared the performance of over forty different feature selection methods implemented in the open-source R community. Provided solution to the supervised learning aspect of the problem with a range of methods including SVM, The Elastic-Net , K-Nearest Neighbors, Quadratic Discriminant Analysis, etc. • Identified, characterized and removed non-biologically relevant technical variability in the data to expose the underlying biological signal through the application of mixed effects modeling and variance components analysis. • Developed custom metrics to quantify spurious biases in the data, and normalization methods to remove these biases. Machine Learning and Bioinformatics Contractor, 12/2011 to 07/2012 Agios Pharmaceuticals – Cambridge, MA • Constructed differential equations based models of enzyme-inhibitor kinetics, and simulated their temporal dynamics with RungeKutta numerical methods. • Employed partial least squares regression to transcriptional data from a collection of cell lines in order to identify the signature of sensitivity to an in-house candidate drug. • Performed meta-analyses combining the results of multiple related transcriptional profiling studies to increase the power to detect transcriptional differences between mutant and wild type cell lines. Senior Research Scientist, Computational Biology, 05/2010 to 09/2011 The Ragon Institute of MGH, MIT, Harvard and The Broad Institute of MIT and Harvard • Collaborated with research and academic professionals to work on the analysis of transcriptional profiling studies primarily aimed at characterizing the responses of leukocyte populations to viral infection. Senior Research Scientist I, 10/2008 to 04/2010 Wyeth/Pfizer • Delivered several outstanding project results as the member of the Systems Biology Analytics group at Wyeth I. • Drove the use of machine learning in the Antibody Stability Project; developed a machine learning based method for the prediction of antibody thermal and acidic stability from primary sequence with the aim of identifying inducible stability improving mutations. • Conducted and published novel research producing new protein sequence embedding methods for machine learning applications. • Supported the Synthetic Lethality Screening Project by providing statistical analysis, visualization, and fitting of dose response data for drug combinations aimed at identifying synthetic lethal interactions among breast cancer treating compounds. • Supported the High Content Screening Project by analyzing the features extracted from digital microscopy data in a high content chemical biology screen to identify common biological activity among a large panel of compounds through the application of factor analysis, unsupervised learning, partial least squares regression, and principal components analysis. • Achieved project deliverables for miRNA Project, including examining the text mining data from pubmed records related to microRNA using term frequency analysis and pathway analysis. Senior Bioinformatician, 01/2008 to 10/2008 Systems Analytics – Waltham, MA • Developed novel proprietary algorithms and methods for the application of machine learning to microarray data and implemented and tested methods for quantifying, visualizing, and removing batch effects from microarray data. • The team successfully ranked as the number one out of the thirty independent teams from academic institutions, governmental agencies, and industrial research centers around the world in the Micro Array Quality Control (MAQC II) Project. The MAQC project was a large-scale collaboration coordinated by the FDA with the goal of developing guidelines for the use of predictive modeling and classification of microarray data. (http://www.fda.gov/nctr/science/centers/toxicoinformatics/maqc/). Research Assistant / Graduate Student, 09/2001 to 05/2007 Boston University, Department of Cognitive and Neural Systems, – Boston, MA • Conducted research for doctoral dissertation on Neural Network and Bioinformatic Designs for Predicting HIV Protease Inhibitor Resistance. Applied several technological components developed to predict the HIV protease inhibitor resistance from viral genotype and identify and analyze those mutations in the HIV protease gene that contribute to protease inhibitor resistance. • Oversaw the development of a new neural network architecture, Analog ARTMAP, for on-line learning of non-linear, multidimensional, and continuous valued maps. Developed an accompanying feature selection wrapper method for post-training analysis of Analog ARTMAP neural networks. • Invented and successfully applied Amino Acid Spaces: a set of general-purpose tools for embedding protein sequences in machine learning applications, combining dimensionality reduction with an interpretable metric derived from Protein Scoring Matrices. Research Assistant / Graduate Student, 09/2000 to 05/2001 Florida Atlantic University, Department of Complex Systems and Brain Science, – Boca Raton, FL • Conducted eye-tracking experiments and research on visual input statistics during free-viewing of natural time-varying images. Teaching Experience Teaching Fellow for CN 520: Principles and Methods of Neural Modeling 2 Boston University, Department of Cognitive and Neural systems, Boston, MA, 2003 Teacher of English as a foreign language, Ichinomiya Minami Junior High School, Ichinomiya, Japan,- Teaching Fellow for PHYS 160: The Physics of Everyday Objects, The University of Michigan, Department of Physics, Ann Arbor, Michigan, 1996 Teacher of English as a foreign language, The Cerro Plano Elementary School, Monteverde, Costa Rica, 1992 Education & Academic Credentials Ph.D., Cognitive and Neural Systems, Boston University, Boston, MA, 09/2001 – 05/2007 • Experienced comprehensive coursework on Machine Learning, Advanced Statistics, Neuromorphic Systems Engineering, Neuroanatomy and Physiology, and Neural Modeling. Complex Systems and Brain Science, Florida Atlantic University, Boca Raton, FL, 09/2000 – 05/2001 • Spent one year in a PhD program before transferring to Boston University. Coursework includes Dynamical Systems and Chaos, Neural Modeling, and Neuroscience. B.S., Double Major in Physics and Mathematics, The University of Michigan, Ann Arbor, MI, 09/1992 – 05/1996 • Completed Bachelor of Science degree with a double major in Physics and Mathematics. The University of Connecticut, Storrs, CT, 09/1990 – 05/1991 • High school student; took advanced evening courses in Ordinary Differential Equations and Linear Algebra at the University of Connecticut. Quinebaug Valley Community College, Danielson, CT, 09/1986 – 05/1987 • Junior high school student; took evening courses in The Mathematics of Finance and College Algebra at the local community college. Technological Acumen • Programming Languages: Python (9 Years), R (9 Years), Bash (5 years), Spark (2 Years), Matlab (12 Years), Java (< 1 Year), Perl (< 1 Year) • Python Libraries: Keras, Tensorflow, Pytorch, Scikit-learn, Xgboost, Dlib, Cv2, Langchain, Chroma, Llama_API, Numpy, Scipy, Pandas, Statsmodels, Joblib, Multiprocessing, Nose, PyInstaller, Flask, Socketio, Eventlet, Matplotlib, Seaborn, Pil, Pickle, Json, Collections, Itertools, Argparse, Datetime, Math, Sys, Os, etc. • Distributed Computing: Spark, some use of Hadoop's file system (HDFS) • Containerization / Encapsulation: Docker (Used in one project to make training portable. Separately used for the inference engine in production), Virtualenv, Anaconda's virtual environments • Cloud Computing: AWS (EC2, S3, Glacier) • Neural Network Architectures: CNNs (ResNet, Retinanet, VGG), LSTM/RNN, Transformers, Others (Back propagating ANNs, ARTMAP, Self-organizing feature maps) • Operating Systems: Mac, Linux, Windows Applied Mathematical Skills Linear Algebra – Ordinary Differential Equations & Boundary Value Problems – Partial Differential Equations – Non-linear Differential Equations – Signal Processing – Time Series Analysis – Multivariate Statistics – Hypothesis Tests – Dimensionality Reduction – Factor Analysis – Supervised Learning – Unsupervised Learning – Splines – Curve Fitting – Regression – Hidden Markov Models – EM Algorithm – Mixed Effects Models – ANOVA – ANCOVA Languages English (Native), Spanish (Fluent), Japanese (Verbal: Advanced, Written: Intermediate) Awards & Recognition • SAIC-Innovation Center Outstanding Service Award, 2018 • Recipient of the Boston University Department of Cognitive and Neural Systems Teaching Fellow of The Year Award, 2003 • University of Michigan Class Honors, 1994 • AFS exchange Student in Ecuador, 1989 to 1990 • Winner of the Pomfret School Olmsted Scholarship, 1988 • Winner of the Marianapolis Scholarship, 1987 • Mathematical Training through CTY (Johns Hopkins University Center for Talented Youth), 1986 to 1989 Patents - Woods M, Wang A, Jiang F, Middleton R. 2018. Personalized Adaptive HVAC System Control Methods and Devices. U.S. Patent Application 16/043,262, Filed July 2018. Patent Pending. - Woods M, Ronca R. 2017. Detection of Anomalous Key Material. U.S. Patent Number-, Filed April 2017. Patent granted July 2020. - Woods M, Ronca R. 2016. Cryptographic Key Control Based on Debasing Condition Likelihood Estimation. U.S. Patent Number-, Filed August 2016. Patent granted January 2020. - Biesinger G, DeBate D G, Nair H R, Ronca R, Woods M. 2016. Assisted Improvement of Security Reliance Scores. U.S. Patent Application-US1. Filed April 2016. Patent granted February 2019. - Bektchiev D, Elarde D, Hill G, Ronca R, Woods M. 2015. Security Reliance Scoring for Cryptographic Material and Processes. U.S. Patent Application-US1. Filed July 2015. Patent granted January 2018 Publications J Judy Chang, Matt Woods, Robert J Lindsay, Erin H Doyle, Morgane Griesbeck, Ellen S Chan, Gregory K Robbins, Ronald J Bosch, Marcus Altfeld, (2013) Higher expression of several interferon-stimulated genes in HIV-1-infected females after adjusting for the level of viral replication. The Journal of Infectious Diseases, doi: 10.1093/infdis/jit262. G Gaiha, K Mckim, M Woods, M Lichterfeld, A Brass, B Walker, (2012) Identification of CD8+ T cell host factors involved in HIV control. Retrovirology, doi: 10.1186/--S2-O44. Woods M*, King A C*, Liu W, Lu Z, Gill D, Krebs M R H, (2011) High-throughput Measurement, Correlation Analysis, and MachineLearning Predictions for pH and Thermal Stabilities of Pfizer-Generated Antibodies. Protein Science, doi: 10.1002/pro.680. *equal contribution. Seyhan A, Varadarajan U, Choe S, Liu Y, McGraw J, Woods M, Murray S, Eckert A, Liu W, Ryan T, (2011) A Genome-Wide RNAi Screen Identifies Novel Targets of Neratinib Sensitivity Leading to Neratinib and Paclitaxel Combination Drug Treatments. Molecular BioSystems, doi: 10.1039/c0mb00294a Rotger M, Dalmau J, Rauch A, McLaren P, Bosinger S, Martinez R, Sandler N G, Roque A, Liebner J, Battegay M, Bernasconi E, Descombes P, Erkizia I, Fellay J, Hirschel B, Miro J M, Palou E, Hoffmann M,Massanella M, Blanco J, Woods M, Guenthard H, de Bakker P, Douek D, Silvestri G, Martinez-Picado J, Telenti A, (2011) Comparative Transcriptomics of Extreme Phenotypes of Human HIV-1 Infection and SIV Infection in Sooty Mangabey and Rhesus Macaque. The Journal of Clinical Investigation, doi: 10.1172/JCI45235. Shi L, et al, (2010) The MicroArray Quality Control (MAQC)-II Study of Common Practices for the Development and Validation of Microarray-Based Predictive Models. Nature Biotechnology, 28(8), 827-38. Vigneault F, Woods M, Buzon M, Li C, Pereyra F, Crosby S D, Rychert J, Church G, Martinez-Picado J, Rosenberg E S, Telenti A, Yu X G, Lichterfeld M, (2010) Transcriptional Profiling of CD4 T Cells Identifies Distinct Subgroups of HIV-1 Elite Controllers. The Journal of Virology, doi: 10.1128/JVI-. Luo J, Schumacher M, Scherer A, Sanoudou D, Megherbi D, Davison T, Shi T, Tong W, Shi L, Hong H, Zhao C, Elloumi F, Shi W, Thomas R, Lin S, Tillinghast G, Liu G, Zhou Y, Herman D, Li Y, Deng Y, Fang H, Bushel P, Woods M, Zhang J, (2010) Enhancing Classifier Prediction Performance by Correcting Batch Effects in Microarray Gene Expression Data. The Pharmacogenomics Journal, 10, 278-291. Murray B S, Choe S E, Woods M, Ryan T E, Liu W, (2010) An In Silico Analysis of MicroRNAs: Mining the miRNAome. Molecular BioSystems, 10(6),-. Woods M, Carpenter G, (2007) Neural Network and Bioinformatic Methods for Predicting HIV-1 Protease Inhibitor Resistance (2007) Doctoral Dissertation. Hobbies & Interests • Certified Rescue Diver • Plays Guitar, and occasionally makes noises on Banjo, & Mandolin • Avid Pickleball and Chess Player