Full-stack Python application that translates YouTube videos into 20+ languages using OpenAI Whisper (speech-to-text), GPT-4.1-mini (translation), and TTS-1-HD (voice synthesis). Real-time WebSocket progress updates, async pipeline, dark mode UI.
Stack: Python, FastAPI, OpenAI API, pydub, FFmpeg, WebSockets, Bootstrap 5
Designed a multi-agent scriptwriting pipeline for educational YouTube content. Five specialized AI agents (Researcher, Writer, Editor, Visual, Branding) coordinate to produce full video scripts from a single topic prompt.
Stack: Multi-agent AI, DeepLearning.AI, OpenAI, Agent Orchestration
Progressive ML coursework from Python data structures through CNNs and transfer learning. Built a SimpleCNN in PyTorch, trained on MNIST, then applied transfer learning to custom handwritten characters. Used GridSearchCV and Optuna for hyperparameter optimization.
Stack: Python, numpy, scikit-learn, PyTorch, Optuna, matplotlib
Analyzed IBM HR Analytics dataset (1,470 employees). Built KNN classifier to predict attrition and linear regression model to predict monthly income. Key finding: OverTime, JobRole, and MaritalStatus are the strongest attrition predictors.
Stack: R, dplyr, ggplot2, caret, KNN, rmarkdown
Python pipeline merging multiple ACS (American Community Survey) Census tables into a unified ZIP-code-level master dataset for Dallas real estate market analysis. Metadata-driven column labeling handles Estimate vs. Margin of Error columns.
Stack: Python, pandas, glob, Census ACS data
Eight-assignment NLP course spanning lexical diversity analysis, tokenization, POS tagging, web scraping (150 IMDb reviews), TF-IDF cosine similarity on Amazon book titles, and VADER sentiment analysis — all in Python with NLTK and scikit-learn.
Stack: Python, NLTK, pandas, numpy, scikit-learn, BeautifulSoup, VADER
Analyzed craft beer market data to identify IBU/ABV trends and propose strategy for underserved regional markets. EDA with ggplot2, geographic segmentation.
Stack: R, ggplot2, dplyr
Simple and Multiple Linear Regression models for Kaggle Ames Housing competition. Built an RShiny web app displaying predicted home prices by neighborhood.
Stack: R, ggplot2, linear regression, RShiny
Interactive educational platform visualizing how misinformation spreads through social networks. Network diagram (nodes sized by influence, links weighted by sharing volume) paired with p5.js generative art simulating the cultural shift from knowledge to superstition.
Stack: JavaScript, p5.js, Network Graphs, HTML5/CSS3
Full-stack automated real estate wholesaling system targeting distressed properties across 7 DFW counties. Continuous operation: public records scraping → lead scoring → multi-channel outreach → contract generation → buyer platform push.
Stack: Python 3.12, FastAPI, SQLAlchemy 2.0 async, APScheduler, ReportLab, Typer, React + Vite
SaaS tool for real estate investors. Upload bank statements and P&Ls; Claude Haiku extracts transactions and computes DSCR; get a lender-ready PDF report with Stripe-gated download.
Stack: Next.js 15, Prisma 5, Clerk, Stripe, Claude Haiku, shadcn/ui
Conversational AI agent that negotiates storage unit pricing using Claude claude-opus-4-6 tool_use API — calls live pricing functions mid-conversation to stay grounded. Sends deal confirmations via Resend.
Stack: Next.js 14, Prisma 5, Claude claude-opus-4-6 (tool_use), Resend
Python backend generating neighborhood intelligence reports from Census ACS and OpenStreetMap data. AI-written prose overview via Claude Haiku; web view and PDF output via WeasyPrint.
Stack: Python, FastAPI, censusdis, overpy, WeasyPrint, SQLModel, Claude Haiku
Content and data platform for real estate professionals — 43+ published articles, custom design system, 6 Gumroad digital products.
Stack: Next.js 15, Tailwind v4, Framer Motion, shadcn/ui, Gumroad API