Projects
Interactive demos embedded below.
Groq AI Playground
An interactive large‑language‑model sandbox built with Streamlit and the Groq API. Users can open multiple chat sessions and test prompts against Groq’s ultra‑low‑latency LLM hardware. The interface streams responses in real time and is ideal for experimenting with complex prompts or assessing model behaviour. The app is hosted on Replit, so it launches quickly and requires no local setup. It uses Python for the backend, Streamlit for the UI, and leverages Groq’s high‑throughput inference to deliver sub‑200 ms response times.
RAGBot
A Retrieval‑Augmented Generation (RAG) chatbot that lets you upload PDFs and text files to create a private knowledge base. After authenticating via Google OAuth, the app ingests your documents into a vector store and uses embeddings to surface relevant passages. The assistant then answers questions with grounded citations, providing “smart search” over your corpus. It supports multiple AI models for natural conversation and runs on a cloud‑hosted backend; the frontend is built with React and Tailwind CSS.
Chatbot with Browser + RAG
A multi‑tool AI assistant that can browse websites, process uploaded documents and even generate/execute code. The home screen lists capabilities such as browsing external URLs, uploading files for RAG‑style Q&A and running Python code snippets. Internally it orchestrates several agents: a headless browser for web scraping, a document processor that embeds text into a vector database, and a code runner. The UI is React‑based and uses websockets to stream chat and code‑execution output in real time.
Agent AI Practice
A Streamlit‑powered sandbox for developing autonomous agents agent-ai-practice-giovanniruiz5.replit.app . Users can spin up multiple chat sessions, select different AI providers (e.g., OpenAI) and customize system prompts or context length. The environment is designed to simulate agent workflows: it exposes internal tools, lets you call them from within the chat and logs each step so you can iterate quickly. It’s hosted on Replit and uses Python/Streamlit on the backend and local state management to preserve conversations agent-ai-practice-giovanniruiz5.replit.app .
Catherine Ruiz’s Website A website for Catherine Ruiz.
A bilingual, mobile‑friendly microsite for an Avon representative. Written from scratch using plain HTML, CSS and JavaScript, it features two pages (“Home” and “About”) with sections in both English and Spanish. The home page displays Catherine’s photo and contact information, including her phone number for orders or beauty consultations. Campaign dates are updated regularly, and the footer credits ruizTechServices for the build.
ruizTechServices|
The business website for ruizTechServices LLC. Built with Next.js and Tailwind CSS and deployed on Vercel, it targets small businesses needing tech advisory and on‑site NYC support. The landing page outlines three core service offerings: on‑site hardware/network support across Manhattan, Brooklyn, Queens and the Bronx; fractional CTO services for architecture guidance; and custom web‑app/AI workflow development. A three‑step “How It Works” section explains the engagement process (clarity call, plan & scope, retainer or project). The site emphasizes direct access to the founder, local presence, no long‑term contracts and async‑first communication.
Chuef.com
A real‑time discussion platform focused on unfiltered conversation. After signing in with Google, users create a Chuef identity and can post to live threads; reading the lobby is possible without an account. The service advertises 24/7 availability, infinite chat length and no posting limits. Built with Next.js and websockets for real‑time updates, it maintains anonymity by keeping Google account details private and stores all data on a backend managed by ruizTechServices. The graffiti‑inspired design underscores its “raw, real, unfiltered” brand.
Ruiz Transformer Hub
A full‑stack platform for training, deploying and managing custom transformer models ruiz-transformer-hub.replit.app . The dashboard displays metrics such as total models, active sessions and API status ruiz-transformer-hub.replit.app . Users can upload datasets, tweak hyperparameters (embedding size, number of layers, learning rate, epochs and dropout) and monitor training progress through interactive charts. Once trained, models can be renamed, downloaded, set as default or deleted via the Model Management section ruiz-transformer-hub.replit.app . The Playground page exposes an API for real‑time inference and returns probabilities and logits, making it ideal for experimenting with new transformer architectures. The app is built with Python/Streamlit and uses PyTorch under the hood.
Nucleus Bot
Nucleus Bot runs entirely on your machine. Ask it to find files, search the web, take notes, control apps — all without sending a single byte to the cloud.