[WIP] An Alexa skill that increases your productivity by answering : What should I do with my free time? Cosmo listens to your Goals, Interests,
and Reminders, and prioritizes them so you have enough time in each day for all of them. Check out more on the Github.
Tertiary classification model created and trained using simple numpy / scipy with data curated and cleaned from Postgres server. Predicts the evaluation
result of a new applicant to the Computer Science House intro process. Extensive testing done with performance of various hypothesis functions, currently
experimenting with effects of regularization and feature scaling.
An independent study with Dr.Homan of the CS Department at RIT to determine which real world events are the most reliable factors for predictive models
to use. We analyse models for their overall performance, and dates when they perform the best and worst. With these insights, and a thorough understanding
of how the models work, we can determine what features are the best predictive measurements. This would ideally culminate in a new ensemble model, or at
the very least advise the public which models are most reliable.
These insights would ideally support political, healthcare, and general lifestyle decisions.
A personal investigation into the effect that collegiate prestige has on monetary success afterwards. This project analyzes salary data from Payscale
associated with major, school, and school type. It is ongoing and has already determined, unsurprisingly, that Ivy League graduates seem to perform
significantly better than graduates from other types of schools. Research currently going into classifying 'prestigious' schools, and determining
what margin of those schools are in the top earning groups.
An LED board turned personal assistant with the aid of a Raspberry Pi, Arduino, and Alexa. Currently automatically updates the date, and there are plans
to integrate Alexa to allow for display of weather and daily reminders. The arduino serves as the main driver for the board, while the pi maintains
functionality for getting any data necessary before rewriting the sketch on the Arduino and reuploading.
Solutions to Andrew Ng's Machine Learning Coursera course. The GitHub repo explains this more in depth, but essentially these are all implementations
of core machine learning concepts such as linear regression, logistic regression, neural networks, backpropagation, etc. There you will find my explanations
of these concepts.
Intelligent Agents : Conversational Project Staffer
Project at MITRE, a Microsoft Bot Framework bot implemented in Microsoft Teams which utilizes Microsoft Graph to intelligently pair employees with
project managers based off of shared interests and other criteria. A project lead starts a conversation with the bot, describes in natural language what
skills they are looking for, availability, etc. The bot reaches out to potential employees with this proposal, then gets back to the project lead with
those who would like the position.
Check it out at hapi-0.appspot.com!
Partner hackathon project made at RIT's Brickhack.
Simple flask app that reads in a handle or hashtag, uses the Twitter API and Tweepy library to pull tweets from that source, then engages
Google Cloud Platform's Sentiment Analysis API to derive a sentiment score on each of those tweets. We serve this information back to the user in an
aggregated fashion by displaying statistics such as their average sentiment, most positive tweet, number of negative tweets, and more.
Never before worked with R, or done the Advent of Code event, so taking this opportunity to become familiar with both. See GitHub for updates on
what I've learned about R as each day goes along.
Group hackathon project made at Hack Dartmouth. Alexa skill which prompts the user for a rap, then creates a response utilizing parts of the user's utterance.
The response is formulated using nltk and a pickled version of a naive bayes classifier, which tags parts of speech. This tagger parses the user's input for
words to use in the response. Then, a context free grammar structures the response and Lambda serves the response back to the end user.
Displayed at RIT's Presidents' Alumni Ball as part of the Computer Science House Smart Dorm Room exhibit. It is a 32x32 LED matrix which is held onto the top
of a desk lamp stand by a 3D printed mount. An Arduino Mega drives the display, which displays one of four designs. A bluetooth module accepts serial
communication to toggle between the various displays and their respective speeds / colors. All designs custom made.
Part of the elevator lobby entertainment system on Computer Science House. Integrated with another house service, HAROLD, such that anytime a member
scans their iButton, this light display turns into a rainbow party. The display is a 150 LED strip of individually addressable LED's, which are driven
by an arduino. A raspi is hooked up to the arduino for remote update abilities.
Project at MITRE, a helpdesk assistant which streamlined employee requests and questions. Accessible by phone or web, employee engages in conversation
driven by Watson Assistant, where they are appropriately redirected to the correct department. Frees up time of helpdesk agents and gets employees to
Discuss mathematical foundations of machine learning and their various applications. Implement the various concepts in Python and / or Matlab.
Subjects studied were : linear and logistic regression, neural networks, superivsed and unsupervised learning, gradient descent, among others.