I am a Machine Learning Engineer & Data Scientist based in New York City with a passion for collaborating with business partners to design machine-learning models and AI-driven solutions that enhance product offerings and empower data-driven decision-making.
What sets me apart is my dual expertise in data science and education. Before transitioning into tech, I was an interdisciplinary teacher, education administrator, and designer with a focus on reimagining mathematics education. My background in educational technology and curriculum development has given me a unique skill set: the ability to craft compelling data stories that resonate with both technical and non-technical stakeholders. I bridge the gap between data and decision-makers, ensuring that everyone—from C-suite executives to frontline teams—can harness the power of data in their work. Whether I’m optimizing an industry classification model, conducting academic research to guide model development, designing interactive tools for client management, or presenting at a conference, my focus is always on creating data solutions that are both impactful and understandable. I currently am getting my Master's in Computer Science with a focus in Machine Learning at Cornell Tech (Cornell University's tech campus in NYC). Previously, I worked as a data scientist at Middesk helping to build/manage machine learning modlels that powered the core engine of Middesk's business identity platform. Prior to that, I worked at BlockFi, where I leveraged cryptocurrency data across the finance, compliance, fraud, and client services verticals to drive business results.
Let’s connect if you’re interested in the intersection of data science, machine learning / AI, education, and business strategy!
Citibike in NYC and other similar bike-sharing systems face a unique challenge in balancing their system. Bikes must be distributed across all stations so that riders have access to both bikes to take out and empty docks to return bikes to. Unchecked, this challenge may cause bikes to pool in a certain station and drain from others. This project attempts to understand which stations in the Citibike system are pools, drains, or balanced. To accomplish this, time series analysis was implemented in Python/Jupyter Notebook to predict the number of bikes at a given station given the time. Then based on their extracted seasonality from the time series model, stations were classified as pools, drains, or balanced using clustering. Lastly, an interactive dashboard was created to provide a system overview of visualizations and station modeling. (Modeling Process Walkthrough | Dashboard Demo)
Kickstarter only recieves revenue when projects only get fully funded by backers. In light of that, this project investigated what factors lead to the complete funding (success) or failure of a Kickstarter project by creating classification models in Python/Jupyter Notebook to predict the success or failure of 191,875 kickstarter projects launched between April 2009 and October 2020. Our final model was able to predict success (a project is fully funded by backers) or failure of a project with 80% accuracy using only information that is would be available to creators at project launch. (Project Walkthrough)
This project attemps to classify images of coronal MRIs of Alzheimers patient's brains into one of 4 progressive stages of dementia using convoluted nerual networks and transfer learning. The goal was to create a model that could be used as a preliminary diagnosis for new patients with early signs of dimentia and limit false negatives. We created several neural networks in Keras/Tensorflow and leveraged transfer learning to improve the recall metric, but ultimately the model was limited by regular brain aging, the overlap between the classes, and the large variance in the non-demented class. (Project Walkthrough)
Towards Data Science
- How to make your own Instagram filter with facial recognition from scratch using python
- How to Collect Live Feed and Frequently Updated Data Using Cron
- Time Series Analysis with Facebook Prophet: How it works and How to use it
- Callbacks, Layouts, & Bootstrap: How to Create Dashboards in Plotly Dash
The Startup
In May 2022, Mitch spoke at the Chainalysis LINKS Conference with colleagues from BlockFi about leveraging the power of blockchain data to generate insights that help drive revenue, increase security and mitigate risk in crypto markets. Check it out on YouTube
- Reinforcement Learning
- Blockchain development
- C++
- iOS Game Dev