e.i.

Hello!

I’m deborah, a miami-based machine learning engineer humanizing AI and solving real-world problems.

Get in touch

👉
👆

BACKGROUND

I recently graduated from the University of Miami with a degree in Computer Science and a minor in Communication. My background spans machine learning engineering, generative AI, analytics, and full-stack development—across healthcare, political, and retail sectors.

I’ve worked on developing, fine-tuning, and evaluating models for both predictive and generative tasks, and I’m especially drawn to tools that translate complex ideas into useful, everyday applications. From building AI-powered workflows to optimizing GenAI prompts, I bridge technical systems and human needs to solve real-world problems with clarity and care.

Outside the lab, I’m a lifelong creative—usually found at the piano, sewing machine, or deep in a book.

SKILLS

FRAMEWORKS

  • Langchain
  • Streamlit
  • FastAPI
  • Express
  • React

TOOLS

  • Git & GitHub
  • VSCode
  • Chrome DevTools
  • Pinecone
  • n8n/Zapier/Make
  • dbt

LIBRARIES

  • scikit-learn
  • PyTorch
  • TensorFlow
  • Pandas
  • NumPy
  • Hugging Face Transformers
  • Matplotlib

LANGUAGES

  • Python
  • SQL
  • JavaScript (ES6)
  • Java
  • HTML / CSS / Sass
  • C++ / C# / C

APPS

ReviewHub

Anatomie Review Hub is an AI fashion platform where users review generated outfits, and feedback feeds custom scoring models that refine prompts, improve future outputs, and guide capsule collection development through automated workflows.

Educational Assistant

Educational AI Assistant is a Pinecone-powered chatbot built for writers, researchers, and students. It ingests uploaded documents to provide fast, accurate, and citation-backed answers, blending retrieval-augmented generation (RAG) with a clean, user-friendly interface.

RECENT PROJECTS

My first portfolio website was a conscious design effort that taught me a lot about HTML, CSS, and information architecture.

  • JavaScript
  • HTML
  • CSS

Retro Haven is a dynamic e-commerce platform built to simulate a modern online shopping experience for college students. Designed as a final project for a Dynamic Data course, the site features programmatically rendered product pages, user-generated content, and JSON-powered data flows. It showcases essentials for student life—from dorm decor to study supplies. The project demonstrates backend data integration, responsive UI design, and real-time updates using structured datasets.

  • JavaScript
  • Node.js
  • HTML
  • CSS
  • Express

Anatomie Review Hub is an AI fashion platform where users review generated outfits, and feedback feeds custom scoring models that refine prompts, improve future outputs, and guide capsule collection development through automated workflows.

  • Python
  • JavaScript
  • Node.js
  • CSS
  • React
  • n8n
  • Express
  • Airtable API

Educational AI Assistant is a Pinecone-powered chatbot built for writers, researchers, and students. It ingests uploaded documents to provide fast, accurate, and citation-backed answers, blending retrieval-augmented generation (RAG) with a clean, user-friendly interface.

  • Python
  • Streamlit
  • Pinecone
  • LangChain
  • gpt-4o
  • psycopg2
  • SQLite

Dianalysis is a data-first nutrition analyzer that converts food labels into calibrated 0–100 diabetes-risk scores, explains the drivers in plain English, and recommends lower-risk same-family swaps. It also includes a dataset builder that pulls and cleans products from Open Food Facts (via barcode), normalizes servings, and groups items (e.g., oats vs. rice) for better recommendations.

  • Python
  • Pandas
  • scikit-learn
  • NumPy
  • Joblib
  • requests
  • Open Food Facts API
  • Jupyter

This tool ingests fashion images and uses BLIP-2 for base captioning, optionally refined with GPT-4 Vision in multi-pass mode (global, construction, pose/lighting). CLIP provides embedding checks for descriptor accuracy, while custom vocab mapping ensures consistent garment terminology. Data and descriptors are managed with FastAPI and Postgres. Built in Python with Hugging Face models, the system exports JSON/CSV descriptors and prompt-ready text lines, and can be previewed with a lightweight Streamlit UI.

  • BLIP-2
  • GPT-4 Vision
  • CLIP
  • FastAPI
  • Python
  • Replicate API
  • Streamlit

LookShift is an AI-powered fashion visualization tool that allows users to see how clothing looks on different body sizes. Upload fashion images and instantly preview them on XS, M, and XL body types using advanced AI image generation.

  • TypeScript
  • CSS
  • HTML
  • Supabase