Christopher Demirjian

Successfully deployed Full Stack website application that acts as a clear representation of my experiences, projects, and credentials.

Interactive Web Application that teaches user how to identify baseball pitch types.

This group project showcases a full stack web application using a Flask framework, HTML, CSS, and Javascript.

CRUD Web Application that users can reference for 2023 Fantasy Football Player information.

This project served as the first iteration of the companion website for the Yale Natural Language Processing (NLP) Lab's paper 'LC2Eval.'

Using the Python Streamlit library, this website allowed for the paper's datasets and findings to be clearly visualized online.

This cloud-based application NFL FanZone is built and designed for NFL fans' casual interaction and NFL-related information lookup, allowing users to query player/coach statistics and view player news.

See PDF for full Project Report.

This program ingests parking ticket data taken from nyc.gov in Parking.csv file, in order to find different statistics of parking tickets

Three smaller programs are implemented:
A stack in Java using an array
A queue in Java using two stacks
A counter in Java that keeps track of the K best objects

This program allows for an interactive game of video poker.
Using Command Line Inputs, the user can choose to keep or trade in cards in their hand, place a bet, and win/lose credits based on the rank of their hand.

Database and associated web application to allow NFL fans to track the statistics, status, and biographical information of each player in the NFL.
The current implementation of this project includes information for all the NFL teams and top players in the 2023 NFL season.

See PDF for full Project Report.

This project trains and evaluates a PropBank-style Semantic Role Labeling (SRL) system by finetuning a pretrained BERT model.
BERT will be used to compute contextualized token represententations and then will be fine-tuned on the SRL task.

This program solves a lexical substitution task (finding correct synonyms given a sentence), using, WordNet, pre-trained Word2Vec embeddings, and BERT.

This model is a trained feed-forward neural network that predicts the transitions of an arc-standard dependency parser

This program builds a Trigram Language Model, which computes three word sequences to determine the probability of this triple of words occurring consecutively.

This program implements a dynamic path planner for a point robot in a 2D terrain discretized as a grid world

Using a Convolution Neural Network, images are able to fed into a model, and output whether the Kuka robot collides with obstacles or not.

Demonstration of Reinforcement Learning.
Man walks around ice grid, avoiding holes to reach goal.
By utilizing reinforcement learning, the man is able to learn how to properly navigate to the goal without falling into holes.

This project contains a variety of programs that use Monte Carlo Simulations to price various call options such as Asian Call Options, Barrier Call Options and European Call Options with the Black Scholes Merton Formula.

This project uses Markowitz Portfolio Optimization to find the optimal weights to maximize returns on given stocks.

See PDF for full projec report.

This project seeks to explore the underlying team statistics that best can predict a team's likelihood of reaching the MLB playoffs.

See PDF for project report.