Portfolio Projects

I designed and implemented a real-time communication bridge between Engineering and Go-To-Market (GTM) teams. By integrating GitHub webhooks with an LLM-powered analysis layer, I automated the translation of raw code commits into commercially readable updates. This system solved the “silent shipping” problem, ensuring Support and CX teams were instantly aware of minor updates without requiring manual input from Product Managers.

Key Achievements:

• Event-Driven Monitoring: Configured Zapier listeners to monitor specific “production-critical” repositories via GitHub webhooks. The system automatically flags Pull Requests (PRs) upon merge, eliminating the need for manual status reporting from engineers.
• Semantic “Plain English” Translation: Deployed an AI agent (utilizing OpenAI/Devin architecture) to parse complex code diffs and commit messages. I engineered the system prompts to strip away technical jargon (e.g., “Refactored API endpoint”) and output value-focused summaries (e.g., “Improved login load times by 20%”).
• Automated GTM Broadcasts: Built a Slack integration that routes these translated summaries directly to relevant CX and Support channels. This created a “live feed” of product changes, reducing internal confusion and support ticket escalation regarding unannounced fixes.
• Operational Efficiency: This pipeline automated the documentation of ~80% of minor patches and bug fixes. This allowed the Product team to shift focus from writing weekly changelogs to high-level strategic release notes for major version launches.

I designed and engineered a “living” data product that predicts Formula 1 qualifying results using historical telemetry data. Unlike static dashboards, this application features a fully autonomous MLOps pipeline that self-updates and re-trains its predictive models weekly without human intervention.

Key Achievements:

• End-to-End Data Engineering: Built a robust ETL pipeline extracting data from the FastF1 API, processing thousands of lap times into a compressed Parquet data store.
• Predictive Modeling: Trained a Gradient Boosting Regressor to predict qualifying grid positions based on Free Practice pace and teammate deltas, achieving a Mean Absolute Error (MAE) of ~3 positions.
• Automated MLOps: Architected a GitHub Actions workflow to run weekly CRON jobs that fetch new race data, retrain the AI model on the latest trends, and auto-deploy the updated application.
• Scenario Analysis: Created an interactive “What-If” simulation tool allowing users to manipulate variables (e.g., “sandbagging” in practice) to test the model’s logic.

The Job Application Tracker is a React-based web app designed to help users organize and manage job applications effectively. Built with Material-UI for a sleek, responsive interface and Firebase for authentication and real-time data storage.

Key features include a real-time search bar, interactive Sankey Chart visualizations (using D3.js) for status progression, and OpenAI integration to analyze and generate cover letters based on user-uploaded resumes.

View Repository

A WebApp that interacts with a Font API to provide searching, trending font lists, and image-to-font detection capabilities. It includes basic interaction functionality allowing designers to preview fonts dynamically.

View Repository

A dynamic trivia generator designed to deliver engaging and interactive quiz experiences. Backend developed with Node.js and hosted on Firebase Functions for serverless processing.

Key Features:

• Randomized question sets and customizable categories.
• Leaderboards and progress tracking.
• Seamless integration with third-party APIs.

View Repository

I played a key role in the end-to-end execution of the Campus division’s major POS system migration. Over an 18-month project lifecycle, we transitioned from an iOS-based point-of-sale system to a scalable Android-based platform leveraging ELO hardware.

I combined technical project management skills with subject matter expertise in campus dining to shape product direction, managing cross-functional collaboration and external hardware partners.

I architected and deployed an autonomous quality assurance agent designed to eliminate the administrative bottleneck of bug reporting. Integrating OpenAI’s LLM capabilities with Zapier workflow automation, this system acts as a first-line triage engineer, enforcing strict quality standards on ticket inputs and autonomously managing the incident lifecycle to protect developer focus.

Key Achievements:

• Automated Quality Scoring: Engineered a custom evaluation prompt using OpenAI to grade ticket completeness against a strict, zero-anchored rubric. The model analyzes critical vectors—client impact, defined scope, expected vs. actual behavior, and reproduction steps—to objectively score incoming issues.
• Conditional Feedback Loops: Implemented a logic gateway where tickets scoring under 75% are automatically reassigned to the reporter. The AI generates context-aware clarifying questions to fill specific gaps, reducing developer investigation loops and increasing ticket resolution throughput by ~10%.
• Semantic Duplicate Detection: Rolled out a Phase 2 upgrade that scans the historical backlog for semantically similar issues. The system autonomously links related tickets and rejects duplicates, reducing backlog noise by ~25% and preventing redundant engineering effort.
• Autonomous Remediation (Beta): Initiated Phase 3 development integrating Devin AI to extend the pipeline from triage to resolution. Designed a workflow where validated descriptions trigger a codebase scan to identify root causes and propose estimated code fixes.

A web-based Yahtzee-style game built with React.js, featuring dynamic scorecards, dice rolling, and live scoring. It uses CSS for a responsive polished UI and ES6+ JavaScript for game logic.

View Repository