
Wesley Welch
Verified Expert in Engineering
Time Series Analysis Developer
After falling in love with statistics in college, Wesley learned Python as a means to facilitate his yearning to use data to answer questions. This has snowballed into his role as a data analyst for an insurance company and a software engineer specializing in machine learning for a risk-management startup. In both roles, he was given a blank slate to create novel solutions to myriad problems, and he sincerely enjoyed it. Wesley thrives in an environment where no two days are exactly the same.
Portfolio
Experience
Availability
Preferred Environment
Windows, Compass, MongoDB, MySQL Workbench, Postman, Jira, Trello, Slack, Visual Studio Code (VS Code), Git, MacOS, Windows 10
The most amazing...
...thing I've developed was an anomaly detection application that compared users to national text/image data from social media sites and public databases.
Work Experience
Machine Learning Engineer
GoodLight, Inc
- Developed a full-stack web application, “Risk EKG,” that gathered data from across the US and allowed users to compare their company’s personnel risk to their industry and location visually.
- Used data-driven approaches to create novel proprietary algorithms that allowed users to calculate a person’s risk across several categories based on publicly available data.
- Designed and executed all phases of ML model creation, from data collection and cleaning, to feature generation, to parameter optimization, to implementation of working models for internal and external use.
- Created a DBSCAN based anomaly detection framework that was feature agnostic and implemented it as an internal API, which allowed employees across the company to perform anomaly detection on a myriad of data sources.
- Developed an LSTM Neural Network for hate speech detection in social media posts, and later expanded the functionality to include sentiment analysis, potential crime detection, and harassment detection.
- Created a convolutional neural network that predicted income for individuals based on their assets (vehicles and properties) and used this prediction compared to actual or estimated income to determine the risk of dubious financial practices.
- Built a Gaussian mixture model that was feature agnostic to perform entity resolution of personal information such as demographics, addresses, and court cases.
- Built natural language processing models trained on context and syntax to identify novel words in social media posts that might be slang or misspelled references to keywords, which gave a more robust keyword detection than using a stored vocabulary.
Founder & CEO
Stellar Futures
- Used Tkinter to and Electron to create a desktop application, “Evolve ACT”, that allowed students to learn topics from the ACT at their own pace, track their improvement, and explore weaknesses to improve performance.
- Built a website landing page using Flask for selling the Evolve ACT product that integrated login capabilities as well as OAuth, utilized a MySQL database, and used Redux to cache user information and progress.
- Built relationships with educational leaders across the southeastern US in order to provide educational enrichment to students for standardized testing.
- Recorded and edited video explanations for every question type and difficulty for ACT test strategies.
Provider Reimbursement Data Analyst
Blue Cross Blue Shield of MS
- Prepared reports and made inferences about client behavior to inform company decisions through the use of SAS and SQL programming.
- Used a combination of actuarial science, statistical analysis, and industry research to determine appropriate prices for novel or updated health care procedures.
- Developed a web crawler using BeautifulSoup and Selenium in order to automatically aggregate research and information about health care procedures and providers.
- Automatically paraphrased lengthy research articles published by Medicare and other government entities into 1-2 page summaries with the use of natural language processing.
- Automated hundreds of manual tasks in Microsoft Excel with Excel VBA and saved my team hundreds of man-hours. One example included taking one column of data thousands of rows long and dividing it equally into 12 columns.
- Created a GUI dashboard with Tkinter for internal company use that aggregated insurance claim data from across the entire state and present clinic comparisons which significantly improved negotiation power with health care providers.
Experience
Crypto Prediction Trading Bot
Evolve ACT
Skills
Libraries/APIs
Flask-RESTful, TensorFlow, Keras, React
Paradigms
Anomaly Detection
Storage
MySQL, MongoDB
Other
Machine Learning, Time Series Analysis, Computer Vision, Web App Deployment, Windows 10, Dashboards
Languages
HTML5, CSS2, Python, JavaScript, SAS, Excel VBA, SCSS
Frameworks
Flask, Compass, Electron, Bootstrap
Tools
Slack, Trello, Jira, Postman, MySQL Workbench, Git, Figma
Platforms
MacOS, Windows, Visual Studio Code (VS Code)
Education
Master's Degree in Computer Science: Data Analysis
Duke University - Durham, NC
Bachelor's Degree in Chemistry
Harvard University - Cambridge, MA