Hello I'm Doron Czarny

Featured Projects

NLP Reddit analysis

NLP Reddit Stress Analysis

Description: Analyzed 147,000+ Reddit posts from 128 U.S. college communities to identify the root causes of student stress. Used a guided classification framework to group highly negative posts into themes like financial issues, academic pressure, and mental health.

Tech Stack: Python, spaCy, scikit-learn, BeautifulSoup, cardiffnlp/twitter-roberta-base

Read the Paper →
ML / AI Research
Atlanta startup dashboard screenshot

Startup Dashboard for the City of Atlanta (Mayor’s Office)

Description: Designed a dynamic dashboard for the City of Atlanta and Emory’s Center for AI to showcase the local startup ecosystem. Features include customizable scatter plots, an interactive heat map, and a searchable companies directory.

Tech Stack: TypeScript, React, Node.js, Express.js, AWS Redshift, DBeaver, Vercel
Note: The database connection has been removed as ownership and hosting were transferred to the City of Atlanta.

View GitHub Repo →
Full-Stack

PennOS (Operating System Simulator)

Description: Implemented a Unix-style operating system simulator, including a priority-based scheduler, a FAT-style file system, and a custom shell. The system supports shell built-ins such as ls and cat, simulates process creation and termination, handles foreground and background execution, and correctly propagates and handles signals within a user-level execution model.

Tech Stack: C, Linux/Unix system calls, spthreads

Systems

Dog Breed Predictor (Image Classification)

Description: Built and optimized a convolutional neural network to classify images of dog breeds. Experimented with dropout, weight decay, batch normalization, and activation functions (ReLU, leaky ReLU, tanh) to improve generalization and reduce overfitting.

Tech Stack: Python, PyTorch, TorchVision, NumPy, Matplotlib, scikit-image

ML / AI

Ghostbusters: Probabilistic Inference in Pacman

Description: Implemented exact and approximate inference (Bayes Nets, particle filters) to track invisible agents in a Pacman variant. Developed belief updates, time elapse prediction, and greedy decision strategies using noisy sensor data.

Tech Stack: Python

ML / AI

Quantum Fourier Transformation Experiments

Description: Simulated and executed quantum circuits to study frequency patterns using the Quantum Fourier Transformation. Compared simulated outcomes to real IBMQ hardware to analyze noise effects.

Tech Stack: Python, Qiskit, IBM Quantum

Research

Medical Insights Dashboard

Description: Built a full-stack healthcare analytics web application that allows users to explore system-wide hospital data, compare insurance providers, and analyze treatment patterns by diagnosis+estimated costs. The application includes authenticated access and supports complex analytical queries across large relational datasets.

Tech Stack: React, TypeScript, Node.js, Express.js, PostgreSQL, SQL
Note: Project received maximum points

Full-Stack

Big Data Analytics on Olist E-Commerce Dataset

Description: Analyzed large-scale Brazilian e-commerce data to study customer behavior, seller performance, delivery reliability, and revenue patterns. Built distributed data pipelines, performed extensive exploratory analysis, and trained machine-learning models to evaluate key business outcomes at scale.

Tech Stack: Apache Spark, PySpark, Spark SQL, MLlib, Python

ML / AI

Unix Shell Interpreter

Description: Engineered a Unix shell in C supporting I/O redirection, background execution, and process management using system calls like fork, execvp, and dup2.

Tech Stack: C, Linux/Unix system calls

Systems

ICU 30-Day Mortality Predictor

Description: Built and tuned a logistic regression model using clinical time-series data. Achieved AUROC of 0.877 and F1 score of 0.512 on test set.

Tech Stack: Python, Pandas, scikit-learn, matplotlib

ML / AI