Risk Mapping: A Hawkes Process Model of Gang-Driven School Gun Violence in Chicago

IPPP & Hawkes Self-Exciting Models. Applied advanced stochastic point process modeling - the same mathematical framework used in algorithmic trading, epidemic forecasting, and cascading event prediction - to quantify how gang environmental density causally drives school gun violence.

Stock Price Prediction via RNN/LSTM Architecture

This project develops a stock price prediction system using a stacked Long Short-Term Memory (LSTM) recurrent neural network built with Keras and TensorFlow. Historical open price data was sourced and a sliding window approach with 60-day lookback period was applied to construct sequential input-out

Pathfinder Algorithm (BFS)

Implemented 50 nodes to perform shortest path calculation on Dartmouth's campus. Visualization accurately reads a file, constructs a graph, and performs various operations on this graph, such as running a Breadth-First Search (BFS), and implementing back chaining to draw the graph efficiently.

Uber Dashboard (SQL, Looker, BigQuery)

Utilized Google BigQuery, MageAI, and Looker after constructing data model containing primary and foreign keys. Utilized 100,000 data points from the Uber rideshare database. Queried columns and fact tables to Google Cloud.

Neural Networks From Scratch

Implimented handwritten digit recognition system using a fully connected feedforward neural network built with Keras and TensorFlow. The architecture consists of an input layer accepting 784-dimensional flattened grayscale image vectors (28×28 pixels), two hidden Dense layers of 128 units

Artificial Intelligence Applications

Artificial intelligence and machine learning models introduce contemporary methods for scientific inquiry. Within this landscape, natural language processing (NLP) techniques enable the analysis of spoken data, while dimensionality reduction methods uncover structures in high dimensional space.

Football Computer Vision Applications (CNN)

Designed and implemented an end-to-end computer vision analytics pipeline to detect, track, and analyze player and ball dynamics in football match footage. Leveraged the YOLO architecture for real-time object detection, identifying players, referees, and the ball.

Origami Classification (CNNs, SVC, Grad CAM)

Trained 3 Convolutional Neural Networks and used Deep learning to classify origami images into 3 difficulty groups (easy:0, intermediate:1, complex:2). The Keras model ultimately achieved a validation accuracy of 0.63 and a 0.79 loss on 1900+ origami images.

LLM Applied Applications (RAG vs CAG)

Users provide natural-language questions, and the system automatically profiles the data, extracts schema-aware metadata, and retrieves only the most relevant context needed to ground SQL synthesis. The design explicitly addresses current RAG limitations including retrieval noise.

Azure Databricks End-to-End

End-to-end Data Engineering applications covering Azure Databricks, PySpark, Spark Structured Streaming, and Delta Live Tables. Created data architecture, ETL pipelines, dimensional modeling, and Slowly Changing Dimensions through real-world use cases and utilized medallion architecture.

End-to-End Azure Databricks Pipeline & Dashboard

Built an end-to-end customer analytics data engineering pipeline on Microsoft Azure to address gaps in customer demographic insights and purchasing behavior. Extracted data from on-premise SQL databases and migrated it into Azure using Azure Data Factory, implementing secure access via Azure Key.

Anti-aging Product Marketing & Media Analytics

This analysis centers on a direct-response TV campaign for an anti-aging skin care product, with three creative spots running simultaneously across multiple stations. The goal was to evaluate campaign performance through the lens of what matters most to the advertiser — cost-per-response (CPR).

Data Structures & Algorithms

K-Means & Hierarchical Clustering

Developed a comprehensive clustering analysis of U.S. utility companies by computing pairwise similarity metrics using Euclidean distance. Preprocessed and standardized multi-dimensional operational data in Python using Pandas and NumPy, then generated a full distance matrices with scikit-learn

Linear & Nonlinear ML Applications (MATH50)

The curriculum integrated key probabilistic concepts, including the Binomial distribution and conditional probability, to frame inference, likelihood estimation, and uncertainty quantification. It also critically examined issues such as p-hacking, highlighting risks of multiple hypothesis testing...

Origami Regression (BERT, XGBoost)

I applied CNNs, decision trees, and regression models (R² = 0.71, 0.817, and 0.582, respectively) to classify origami structures. Additionally, I managed to develop a Topic Model (k = 7) and decision trees (R² = 0.69, boosted) to predict origami complexity based on fold time.

Weather Forecasting (ARIMA)

I analyze the Jena Climate dataset, which contains weather mea- surements recorded every 10 minutes from 2009 to 2016. The analysis focuses on uncovering daily and monthly weather patterns through data preprocessing, aggrega- tion, reshaping, and forecasting and visualization.

Advanced Sales Analytics (Power BI)

Built a professional Coca-Cola Product Analysis Dashboard in Power BI, showcasing end-to-end data visualization skills—from data cleaning and modeling to DAX-driven KPIs and interactive visuals. The dashboard highlights sales, profit, and product performance across regions, with dynamic filters.

Data Quality Assessment (Lead Generation)

This data cleaning exercise and recommendation task prepared a prospect database for immediate sales outreach in Hubspot. A data quality assessment identified 5 top issues: incomplete data, expired data, incomplete site details, ambiguous program classifications, and a lacking data dictionary..

Crunchyroll Dashboard (PowerBI)

Dashboard visualizes the Crunchyroll anime database by highlighting key performance indicators over 1,200 titles. Each anime entry includes its English name, image, number of episodes, total votes, etc. With an anime average rating of 4.08, the dashboard highlights audience trends on Crunchyroll.

NETFLIX

Engineered an ELT data pipeline integrating DBT, Snowflake, and S3, implementing data quality testing and designing data architecture for scalable ingestion and transformation

SpaceX Operational Analytics (Starlink, Falcon 9)

Collected Starlink satellite and Falcon 9 launch data (2010-2022) and visualized a compelling infographic in three.js. Imported two 3d models (GLTF) in JavaScript. Utilized KPIs and launch data specifically in regard to Starlink satellite efficiency based on a working to deployed satellite ratio.

Astrophysics Research & Visualizations (ASTR25)

Studied galaxy formation and dynamics, clustering, dark matter, and large-scale structure; explored Big Bang cosmology, including FRW equation, nucleosynthesis, and cosmic microwave background.

Speed Records (RStudio, Data Reporting)

Analyzed the progression of production car speed records using R, focusing on key milestones such as the Mercedes-Benz 300SL (~200 km/h in the 1970s), the Porsche 959 (305 km/h in 1986), and the Koenigsegg Agera RS (447 km/h in 2017). Utilized regression models and visualization libraries.

Housing EDA (Snowflake)

Loaded housing data. Achieved key insights. Developed a correlation matrix/ heatmap in Snowflake by leveraging Cortex Code and Agents. Provided dependency graph as reference. Migrated data from SQL to Python to visualize results

Hotel Booking Customer Analytics (Snowflake, SQL)

I designed and implemented a production-style data pipeline in Snowflake using the Medallion Architecture (Bronze → Silver → Gold) to process raw hotel booking data end-to-end. I developed fully structured, analytics-ready datasets and created an interactive dashboard in Snowsight.

Evolutionary Game Theory Model of the Renewable Energy Market

This study explores the factors that influence consumer choices on whether to invest in renewable energy sources versus fossil fuels, focusing on solar and wind energy. Motivated by the current climate crisis, highlighted by the IPCC’s projection of a critical global temperature rise by 2030, and th

Interactive Chess Visualization (EDA)

Implemented an interactive UI feature to a statistical representation of my rating progress on the website chess.com. Scraped some of the data in python and utilized APIs to finalize the visualization based on my rapid, blitz and bullet ratings from the year 2017- 2022.

Kino SMPL Model (Sage Maker, Version Control)

The SMPL (Skinned Multi-Person Linear) model is a statistical 3D human body model that represents the surface of the human body with a triangular mesh of 6,890 vertices and 13,776 faces, deformed by both pose and shape parameters.

COVID-19 Research Paper (Qualtrics, SurveyMonkey)

Research study looks to highlight the nuanced role of how political partisanship influences engagement in a compliance to take the COVID-19 vaccine. Created sectioned groups that either had or didn't have a vignette. The vignette was intended to increase compliance to take the coronavirus vaccine.

Spotify Radar Plot

D3.js radar plot visualizing six music attributes (e.g., danceability, liveness, energy) with hover and highlighting features for quick classification. Interactive controls include search, dropbox, checkbox, and reset, enabling exploration by artist, year, or genre in a single view.

Ori(go) Web App

Ori(go) is a social media platform and digitial repository where users can collect, augment, and trade a plethora of origami models based on their real world creations. After taking a picture of origami, user models are digitized then transferred into an online origami repository.

Ori(go) (YCombinator: Summer Application, Ideation Phase)

Orig(go) is a long term business project that involves origami in the digitial realm. Origami is the mathematical deconstruction of a 2d object to the 3d realm. Ori(go) is based on origami inspired by Transformers and LEGO; features a repository collection in the epic style of Pokémon.

USA Shootings Viz (R: Patchwork)

Created a visual representation illustrating the frequency of shootings in the top three most populous U.S. states, followed by an analysis of their respective lethality. Studied contributing factors to the shootings (whether it may be the population or gunsales affecting the number of shootings).

Pizza Sales Dashboard (SQL, KPIs)

Leveraged Excel, SQL and Microsoft Power BI to analyze key metrics in pizza sales. Calculated total revenue, average order value, total pizzas sold, total orders, and average pizzas per order. The dashboard features bar charts, line graphs, donut charts, and utilizes KPIs and DAX functionalities.

NBA Player vs Team Efficiency

Project based on the contribution of members in Data Sports Analytics to determine the efficiency level of major NBA teams based on whether individual efficiency (teams with 1 major elite player) trumps holistic efficiency (teams with a collective of average players)

The Rise of Magnus Carlsen

By 2010, Carlsen had already reached World #1 in the FIDE rankings, becoming the youngest player to ever hold that position at the time. His Elo rating peaked at 2882 in 2014, the highest ever recorded in history, marking a statistical dominance unmatched even by previous champions like Garry Kaspar

Animated US Shootings Plot

Visualized and animated USA shootings data illustrating the frequency of shootings in the top three most populous U.S. states, followed by an analysis of their respective lethality. Studied the uprise of mass shootings primarily after the 2010s.

The Influence of Gang Presence on School Shootings

California has the highest number of school shootings and despite having stringent gun laws, the state has experienced the most school shootings in the US. I will perform a comparative analysis and compare how a state with a low gang population compares to California ( a state with a high gang pop)

Dartmouth Medians (Interactive, DALI)

An interactive D3 plot highlighting the median GPA and enrollments across Dartmouth College departments. Incorporates hover features, drop box, radio toggle features, and separates Department based on internal classifications

DALI Median & Superstore Projects

Basic representations showcasing the distribution of dartmouth medians. Assembled distribution plots, assesed the difficulty of a major based on grades of Dartmouth students, inquired whether larger classes equated to lower grades and deciphered whether the difficulty of a class is major-specific.

Bio Electric Vehicles (BEV) Heatmap

Visualized data relating income level to the likelihood of owning a bioelectric vehicle (BEV). In recent years, electric cars have been a topic of conversation but even though BEV vehicles are more expensive upfront... are they more preferred by the rich or those in urban neighborhoods?

The Unexpected Rise of Pickleball

Being a relatively old sport from the 1960s, pickleball has taken off and has experienced a renaissance while also containing major aspects of tennis, table tennis, and squash. I am interested in the causal mechanisms at play that has resulted in the sport becoming so popular

The Gender Wage Gap (EDA)

The gender wage gap refers to the difference in average earnings between men and women. It’s usually expressed as a percentage of men’s earnings—for example, if women earn 82 cents for every dollar men earn, the wage gap is 18%.

Ori(go) Investor Pitch (Market Entry)

Investor Pitch (Market Research)