Data Scientist
Resume:
Jason Ngo
Sunnyvale, CA *****
add6kh@r.postjobfree.com
OBJECTIVE
Looking for a machine learning/data science position EDUCATION
UC San Diego — BS in Mathematics and Computer Science, Class of 2020 PROFILE
I am an undergraduate research assistant working in the intersection of machine learning, computational geometry and mathematical neuroscience. In the industry, I have used Deep Learning to solve natural language processing problems. I have also developed algorithms to automate data analysis of a reactor. My strengths include independent thought, collaboration, consistency and focus.
LANGUAGE AND SKILLS
Java, Python, Ruby on Rails, C++, MySQL, Javascript Keras, Tensorflow, PyTorch, Scikit-learn, spaCy, Hadoop, PySpark, Hive, Deep Learning, Convolutional Neural Networks, Natural Language Processing, Data Analysis Automation, 2-D Material Synthesis, Machine Learning Architecture, Computational Geometry, Random Geometric Graph Theory
EXPERIENCE
Data Science Intern @ American Express
New York, NY June 2019—August 2019
Supervisor: Alireza Aliamiri. Available upon request Member of the GCS AI COE (Global Commercial Service Artificial Intelligence Center of Excellence) team. My project was the AP File Standardization. Our objective was to map the contents of an arbitrary customer’s database file to 20 categories in the American Express’ template. My responsibility was to create a model that predicts the label given a column’s header and contents. This project replaced a 3rd party vendor who charged $900 per file.
• Initially my job was to modify a word2vec architecture that maps each column header to a vector in R300, which is then combined into the feature space of a random forest model. This improved the performance of our random forest model by 8-10% for 4 categories, resulting in the model’s overall performance of 82%.
• In an effort to improve the robustness of my model, I proposed and implemented a new architecture for a character embedded convolutional neural network (CNN) using 4 convolutional layers of varying sizes followed by 3 multi-perceptron layers and an output layer. By using a character embedded CNN, the new model was not only robust, but also performant. It handled edge cases such as misspelled or previously unseen headers and improved the performance for the classification of 20 categories to 96%. As a result, it was selected for further development and production. Machine Learning Undergraduate Research Assistant at CENI@UCSD La Jolla, CA, March 2018—Present
Advisor: Gabriel Silva. Available upon request
We are collaborating with Microsoft Research on the dynamic perceptron project to develop new machine learning architecture. Our objective is to develop a metric to compare the dynamics of two neural networks. My work includes
• Developed an algorithm that projects a temporal graph onto a geometric graph in R3. This algorithm was presented at the CENI Symposium 2018.
• Computed percolation threshold, critical connectivity and clustering coefficient of this transformed graph.
• Currently developing dynamic graphlet algorithms to analyze the topology of the network.
Energy Intern @ SRI International
Menlo Park, CA, April 2016 — September 2016
Advisor: Francis Tanzella. Available upon request
Our team collaborated with Brillouin Energy and the Anthropocene Institute to study CECR (Controlled Electron Capture Reaction) using the HHT (Hydrogen Hot Tube) reactor.
• Initially my job was to manually analyze data of the HHT reactor. Upon noticing the patterns in the data, I began to develop an algorithm to automate the data analysis. This algorithm was implemented in Java. Previously, it took an independent data analyst and an SRI scientist at least 2 weeks to do data analysis on a month of data from the HHT reactor in order to determine the excess power. My program completed this task in just 5 minutes
Intern in MXENE Group @ Drexel University,
Philadelphia, PA — June 2014 — August 2014
Advisor: Michel Barsoum. Available upon request
After working on a Beryllium battery project in my junior year of high school, I was fortunate to get into Dr. Barsoum’s MXENE lab. There, I developed battery anodes using different MXENEs, a 2-D material, and performed exclusion zone water experiments using MXENEs to study the properties of super-capacitors. CLASSES
Abstract Algebra (Math 100A, 100B and 100C)
Real Analysis (Math 140A and 140B)
Statistics (Math 181A and 181B)
Combinatorics (Math 184A)
Data Analysis and Inference (Math 189)
Number Theory (Math 104A)
Computer Systems and Organization (CSE 30)
Advanced Data Structure (CSE 100)
Design and Analysis of Algorithms (CSE 101)
Theory of Computability (CSE 105)
Introduction to Machine Learning (CSE 151A)
Deep Learning (CSE 154)
Contact this candidate