Vasari · The Battle of Marciano · 1572
Work

Selected projects

Six projects across AI, software, data, and social platforms.

Wysdom AI — mobile study feed of active-recall questions

Wysdom AI

Wysdom AI is an AI-powered study platform that turns passive studying into active recall. Students upload class notes, lecture slides, or study guides, and the app converts that material into a personalized, infinite vertical feed of questions — a TikTok-style feed built on proven retrieval-practice techniques for long-term retention. Learners get immediate feedback, revisit weak concepts, and build mastery through continuous active recall.

Built with React Native and Expo for a single codebase across iOS, Android, and web. Uploaded documents are parsed, cleaned, and segmented, then passed through an AI pipeline that summarizes the material, extracts key concepts, maps relationships between topics, and generates multiple-choice questions, flashcards, and open-response prompts. A cloud backend handles progress, analytics, and authentication, while AI inference is tuned to minimize latency and cost.

How it works
  1. Upload your notes, lecture slides, or study guides.
  2. The AI pipeline summarizes the material and generates multiple-choice questions, flashcards, and open-response prompts.
  3. Study through an infinite feed of active-recall questions with immediate feedback and mastery tracking.
View on the App Store →
Signal — an SEO and GEO readiness audit gauge

Signal Audit

Signal is a browser-based audit tool that scores any website for both classic search (SEO) and AI answer engines (GEO). It crawls up to 60 internal pages, extracts on-page signals, and produces a deterministic 0–100 score across each dimension — the same input yielding the same result every run.

For SEO it measures titles, meta descriptions, heading hierarchy, word count, alt-text, structured data, and links. For GEO it evaluates schema, answer-ready content, fact density, extractable formatting, authority, entity clarity, and AI-crawler access. Google's Gemini API generates an action plan — executive summary, quick wins, answer-engine simulations, and prioritized fixes — and Signal auto-generates robots.txt and llms.txt. It runs entirely client-side with no backend, deployed on GitHub Pages.

What it checks
  1. SEO signals: metadata, headings, content depth, structured data, media, and technical health.
  2. GEO dimensions: schema, answer-ready content, fact density, extractable formatting, authority, entity clarity, and crawler access.
  3. Generates an AI action plan plus ready-to-ship robots.txt and llms.txt.
Open Signal →
Social media network of high-growth Instagram accounts with an organic growth curve

Social Media Agency

A portfolio of high-growth Instagram accounts totaling 15M+ followers — including @ocean.destinations (5M+) and several other 1M+ pages — generating 100M+ organic impressions each month.

Over 8+ years I've built and scaled these communities through content strategy, audience analytics, and trend forecasting; developed automated content strategies and publishing schedules; analyzed performance metrics; and optimized engagement with data-driven decisions. I've led designers, video editors, and developers to produce high-performing content, and grown multiple accounts from the ground up to 1M+ followers across several niches.

Highlights
  1. 15M+ followers across the network, including @ocean.destinations (5M+) and multiple 1M+ accounts.
  2. 100M+ monthly organic impressions built over 8+ years through content strategy and trend forecasting.
  3. Automated content strategies, publishing schedules, and data-driven engagement optimization.
  4. Grew multiple accounts from zero to 1M+ followers across several niches.
View on GitHub →
Formula Hybrid optimization — regression scatter with a fitted line and a marked optimum

Formula Hybrid Optimization

An R-based study of Formula Hybrid and FSAE endurance data that models total scoring as a function of lap time and energy usage. Two linear models — one for efficiency, one for endurance — are combined into a single objective, then numerically optimized to find the lap time and energy draw that maximize total points.

The endurance model achieved R² = 0.988; the efficiency model was noisier but directionally clear. Useful as a planning tool for the team when deciding how to trade outright pace against energy budget.

To read
  1. Download the PDF and open in any reader. All code, output, and plots are inline.
Download PDF ↓
AI Chess — a board with a knight's L-move and a minimax decision branch

AI Chess

A full chess engine built from scratch for CMU's 15-112 course. Handles the complete rule set — castling, en passant, promotion, check, checkmate, and stalemate — with an AI opponent that prioritizes capturing high-value pieces. Built on cmu_graphics with a menu-driven interface.

To run
  1. Install the cmu_graphics package and place it alongside the source.
  2. Launch Main_Menu.py to start.
  3. Return to the main menu at any point from within the app.
View on GitHub →
Social Signals — speech bubbles and a sentiment distribution of political posts

Social Signals

A Python-based analysis of bias and emotion in political Facebook and Twitter posts. Uses pandas for cleaning, NLTK for sentiment scoring, and matplotlib for visualizing trends by hashtag and account type.

To run
  1. Install dependencies: pandas, matplotlib, nltk.
  2. Run hw6_social.py.
  3. Call runWeek1(), runWeek2(), etc. for modular analysis.
View on GitHub →