Jaime Cabrera

Rockstar Games Datathon

• Selected as one of 40 UC Berkeley undergraduates to participate out of a competitive pool.
• Analyzed three months of gameplay data from 9,476 Grand Theft Auto Online players across Xbox, PC, and PS4.
• Implemented a K-Means clustering algorithm to investigate player behavior, preferences, and lucrative trends.
• Delivered insight-driven targeting strategies for player engagement and in-game content development for sustained revenue growth.

Ham/ Spam Email Classifier

• Trained a binary classification model based on the Apache SpamAssassin database of emails.
• Employed Python libraries including Numpy, Pandas, Seaborn, Scipy, SKlearn, and Matplotlib to train a model that detected spam emails.
• Model achieved over 93% validation accuracy.
• Performed feature engineering, exploratory data analysis, and logistic regression.

Stratify

• A version of Spotify's Wrapped, an application that enables Spotify users to browse through their favorite artists and track their listening activity, like top tracks, artists, and other stats year-round. Developed a seperate analytics page that displayed and created personal reflections of users based on their listening patterns.
• Created a recommended playlist based on users top songs, artists, and genres using Spotify API.
• Employed Implicit Grant method authorization, to get user approval to access data through the Spotify Web Developer API.

Gitlet: version control system

• Implemented a version-control system for local and remote repositories using Java.
• System mimicked major features of Git, including commit, branch, merge, push, and pull.
• Designed persistence employing serialization and a Cryptographic hash function (Secure Hash Algorithm 1) to store and reference meta data.

VIDEO GAME: 2048

• Built my own version of the iconic 2048: sliding tile puzzle game.
• Implemented identical features to the original game. Allowed players to choose the tilt direction via keyboard arrow keys, player earns points through merges (maximum score is stored in persistent memory and updated accordingly).
• Applied the core game logic and implemented the tilt operation using two software design patterns: the Model-View-Controller Pattern (MVC) and Observer Pattern.

MOVIE GENRE CLASSIFIER (MACHINE LEARNING, NATUAL LANGUAGE PROCESSING)

• Constructed a k-NN classifier that classified movies into one of 5 genres based on the frequency of various words in a movie script, using a dataset from Cornell University.
• Employed feature engineering techniques with linear regression to select features from a bag-of-words to construct a supervised learning model.
• Model achieved 89% test-set accuracy.
• Built a separate classifier using logistic regression.

CARDIOVASCULAR DISEASE: CAUSES, TREATMENT, AND PREVENTION

• Investigated the major cause of death in the world, looking at decades of medical research (dated 1900-2015).
• Analyzed multiple causes and effects of cardiovascular disease across four different studies.
• Examined the risks, treatments, and preventions through statistical hypothesis testing (p-values, simulated test statitics), relative risk probability, exploratory data analysis, and A/B testing.
• Concluded that a low saturated fat diet caused a significant difference in the death rate of participants in the National Heart-Diet Study.

Software Engineering Intern

JPMorgan Chase & Co. (NYSE: JPM) is a leading global financial services firm with assets of $2.6 trillion and operations worldwide. The Firm is a leader in investment banking, financial services for consumers and small businesses, commercial banking, financial transaction processing, and asset management. A component of the Dow Jones Industrial Average, JPMorgan Chase & Co. serves millions of consumers in the United States and many of the world's most prominent corporate, institutional and government clients under its J.P. Morgan and Chase brands. Information about JPMorgan Chase & Co. is available at www.jpmorganchase.com.

• Built SVM and Naive Bayes model to classify Public Cloud SRE incident reports (Natural Language Processing).
• Improved operational efficiency and accuracy in incident classification, reducing classification time by 80%.
• Implemented new support chatbot using Llama 2 LLM and LangChain for automated incident handling.
• Provided leadership, training, and detailed documentation to offshore tech team in project knowledge transfer.

Software Engineering Fellow

• Utilized business logic to develop software that helped solve issues in under-banked communities.
• Led team to promote growth and scheme out digital expansion in a minority depository institution's customer base with targeted marketing strategies, and presented business case to C-suite executives.
• Ideated and developed hackathon website to protect sensitive banking information using MERN stack.
• Placed first in an internal robotics competition, among other tech locations within the firm.
• Attended variety of workshops to enhance and gain skills in business case development, Microsoft Excel, and project management.

Project Volunteer

Project partnered with the Downtown San Diego Partnership that aimed to digitize and analyze hand-recorded homelessness counts. Involved geo-referencing five-years worth of monthly maps, to produce a single digital geographic dataset in ArcGIS, which was later analyzed for time trends and for the association between homeless movements, geography, and the built environment.