This is where I currently work. My work here has varied quite a bit but a summary is as follows:

- Pioneer member of data team
- Initial development of the data stack (date warehousing, reporting database, ETL, development, productionizing)
- Initial growth of the data team including hiring, mentoring, OKR setting, and software implementation
- Cross-functional work with customer experience, development, and executive teams
- Company wide financial and customer forecasting (ARIMA, Prophet)
- Applications of many NLP methods for brand/creator matching, similarity scores, topic modelling (methods include using    Word2Vec, Doc2Vec, GloVe, various Hugging Face models including BERT, GPT, and OpenAI's API)
- Development of private Python libraries and docker containers for model deployment
- Deployment through Sagemaker & Flask
- RShiny and Flask applications for internal machine learning servicing
- Applications of custom Reforge methods to identify and isolate user behavior (setup, aha, habit moments)
- Primary source for validation on A/B testing and experimental design
- Modelling: NLP (word2vec, LDA, etc.), Supervised (regression, tree methods, neural networks, etc.), Unsupervised (PCA, clustering methods, etc.), Monte Carlo Simulations, etc.
- Programming: R, Python, SQL
- Software: AWS (SageMaker, S3, ECR, EC2, Lambda, etc.), Amplitude, Big Query, Google Compute Engine, Fivetran, Segment, RShiny, Flask, Google Analytics, Notion, Zendesk, Asana

At the moment, I'm working on some new product features.

Abstract: We introduce a methodology for integrating functional data into densely connected feed-forward neural networks. The model is defined for scalar responses with at least one functional covariate and some number of scalar covariates. A by-product of the method is a set of functional parameters that are dynamic to the learning process which leads to interpretability. The model is shown to perform well in a number of contexts including prediction of new data and recovery of the true underlying coefficient function; these results were confirmed through cross-validations and simulation studies. A collection of useful functions are built on top of the Keras/Tensorflow architecture allowing for general use of the approach. Note: previously titled 'Functional Neural Networks'.

Check out the latest version at AMSTAT.

Abstract: Neural networks have excelled at regression and classification problems when the input space consists of scalar variables. As a result of this proficiency, several popular packages have been developed that allow users to easily fit these kinds of models. However, the methodology has excluded the use of functional covariates and to date, there exists no software that allows users to build deep learning models with this generalized input space. To the best of our knowledge, the functional neural network (FuncNN) library is the first such package in any programming language; the library has been developed for R and is built on top of the keras architecture. Throughout this paper, several functions are introduced that provide users an avenue to easily build models, generate predictions, and run cross-validations. A summary of the underlying methodology is also presented. The ultimate contribution is a package that provides a set of general modelling and diagnostic tools for data problems in which there exist both functional and scalar covariates.

Check out the paper here.

You can also take a look at the website for the package here.

This was a mammoth solo effort. I created the website from scratch using a React frontend. The website covers everything from the foundations of NLP (namely Linear Algebra, Calculus, & Statistics) to modern approaches including detailed explanations on much of the math underlying models like GPT and BERT. Many other methods are covered as well including CNNs, RNNs, Seq2Seq structures, Autoencoders, etc. The website stands as a public educational resource.

NLP Begin.

Another solo effort. I created a web application using a React frontend paired with a Ruby backend. The backend framework was Sinatra. The ML deployments were done through Flask. The website allowed users to register, login, and use various financial tools to gain insights on securities. The web app was served through AWS and much of the specific ML model APIs were exposed through Elastic Beanstalk. A much more detailed outline is provided in the link below.

AstroML.

An extension of functional neural networks for when the response is infinite dimensional. These types of networks will be useful for when the objective is to predict densities or curves. This extension exists in a form as a part of the FuncNN package but is subject to a full implementation and paper coming soon, maybe.

A gentle introduction to the basics of Gaussian Processes. I introduce the basics from a Bayesian perspective using the canonical example of Bayesian linear regression. Some R code is provided at the end of the paper that shows the readers how to implement this approach without the use of a package.

Project.

This project won second at the annual case study competition held by the Statistical Society of Canada. We used a unique combination of propensity scores and LASSO regularization to assess the relationship between heart disease and Osteoarthritis.

Project.

This is an introduction which includes the foundation for this view of neural networks. The authors (Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, David Duvenaud) had the insight that the usual neural network approach was essentially using Euler's method to work through its layers. They take a step back in this process and move forward in a different direction using more sophisticated approaches that, in some sense, allow for an infinite number of layers. Caveat: this is a bit of an older project.

Project.

A presentation on an extension to functional single index models that has the added useful property of shrinking down the continuum domain to its relevant subset. The working paper is done by my supervisor Jiguo Cao, and has been submitted for publication.

Project.

Least-Angle Regression (LARS) is an algorithm for fitting linear regression models to high-dimensional data, developed by Bradley Efron, Trevor Hastie, Iain Johnstone and Robert Tibshirani. This project re-presents the mathematical details outlined in the original paper in a more approachable manner.

Project.

I sat for and passed this exam on April 17, 2022. I had studied for it in random bursts during the couple months leading up to it using a number of resources, most of which I have listed below! In general, I found the exam to be OK; there were a number of questions on there that were not really covered in any of the study material and most of these questions were related to new AWS feature releases. In particular, things like Model Monitor, AWS Data Wrangler, AWS Sagemaker Studio, BERT, Shapley, and some new S3 functionalities showed up that I wasn't really prepared for. Also, it would probably have helped to have a lower level certification that covered the basics of AWS since much of the challenge of this certification is just trying to remember details about the large number of topics it covers -- these being split between generic ML things and specific AWS services. Anyway, here are some resources:

Course(s):
1. https://www.udemy.com/course/aws-machine-learning/
2. https://explore.skillbuilder.aws/learn/course/27/play/54/exam-readiness-aws-certified-machine-learning-specialty
3. https://sudo-code7.github.io/mls-c01/assets/data/AWS-SM-Built-in-Algos-Compared.pdf
‍
Practice Exam(s):
1. https://www.udemy.com/course/aws-certified-machine-learning-specialty-full-practice-exams/
2. https://www.udemy.com/course/aws-machine-learning-practice-exam/
3. https://portal.tutorialsdojo.com/courses/aws-certified-machine-learning-specialty-practice-exams/
4. https://d1.awsstatic.com/training-and-certification/docs-ml/AWS-Certified-Machine-Learning-Specialty_Sample-Questions.pdf
5. https://amazonmr.au1.qualtrics.com/jfe/form/SV_2nt0z0Far6DzAwJ

Good luck!

This is another article that is a more friendly reading of the paper 'Deep Learning with Functional Inputs'. I go over neural networks and the basics of FDA and then consolidate the two. Take a look here: link.

This is an article I wrote a couple years back on functional data analysis (FDA); it's an introduction that goes over the basics of FDA including basis functions and functional linear regression. You can read the full article over here: link.‍