About Adam
A History of Technical Excellence
Adam is a Data Engineer with 4 years experience in a technical role. He is currenlty working with ETL Pipelines, data warehousing and automation within Asset Management. His career experience has been multi-faceted including areas such as Data Analytics, Business Analysis, QA, Data Science, Machine Learning and more. No matter how foreign the task, Adam has stepped up and completed it with great success and without hesitation.
Dedication to Continous Education
Adam graduated Magnum Cum Laude with a degree in Industrial Engineering from Arizona State University. Since then he has continuously taken courses on Udemy and used those skills to work on software engineering and machine learrning projects. Adam not only has a thirst for knowledge he has a desire to teach. He was a teaching assistant for a coding bootcamp where he mentored students on how to use code for data analytics.
Learn More →Web Development
HTML
CSS/SASS
JavaScript
ReactJS
AWS
Python Development
OOP
Flask
SQLAlchemy
Pandas
Sci-kit Learn
Data Engineering
SQL Server
PostgresSQL
Data Modeling
NoSQL
Spark
Machine Learning
Natural Language Processing
Tensorflow
Outlier Detection
Predictive Analytics
Classifcation Models
Main Projects
Haiku Generator
- Python
- TensorFlow
- RNN
- BERT
- GPT-2
In this project I use multiple NLP techniques to generate haikus from a given seed
Variable Encoder
- Python
- Data Preprocessing
- Logistic Regression
- TensorFlow
- XGBoost
This project explored a new method for encoding categorical varables by assigning each variable a ranked number based on how it impacts the target
Baby Name Generator
- JavaScript
- HTML/CSS
- AWS
- Python
Application that will recommend a name for your baby based on a few simple questions
Haiku Generator
Source Code
The goal of this project was to come up with a method for creating Haikus with a 5-7-5 syllable structure. The first method attemtped was to try to use an LSTM in TensorFlow. This was done by transforming multiple sets of haiku data into phonemes using an enlgish to phoneme dictionary provided by Carnegie Mellon. A method for converting from a phoneme back to an english word was developed using BERT to predict a phoneme from a list of homophones for example: to, too, two. After this was completed the homophones were used to create an RNN model in TensorFlow. Unforunately the poems the model produced were mostly incomprehinsible and did not always follow the 5-7-5 sturcutre. Because of this the haikus were instead used to build a model using GPT-2.
View ProjectCategorical Variable Encoder
Source Code
The purpose of this project was to create a module that could encode categorical varialbles better than both One-Hot Encoding and Label Encoding. The previous method both had considerable drawbacks which I have always found frustrating. Label Encoding is able to create a single column but also creates an order between variables where there isn't one. One-Hot Encoding does not have the issue of creating a false order but it has the potential to create hundreds of columns if there are too many variables. To solve this issue, I developed a module that would create an ordinal column which is correlated to the target variable. This way we can hold the variable to one column but the order is no longer arbitrary.
View Project