maria c. lima

Exploring the possibilities of artificial intelligence through practical applications.

AI Projects

Stock Agent

AI-powered stock analysis dashboard with real-time data visualization and on-demand insights. Features FinBERT sentiment analysis for news articles with 70-90%+ confidence, RAG (Retrieval-Augmented Generation) with ChromaDB vector store for contextual insights, and OpenAI GPT-4o-mini for intelligent stock analysis. Implements batch processing for performance optimization, achieving sub-3-second initial load times. Supports US stocks and cryptocurrency with Polygon.io API integration, interactive charts, and comprehensive financial metrics.

Next.jsReactTypeScriptFastAPIPythonOpenAI GPT-4o-miniFinBERTLangChainChromaDBRAGPydanticPolygon.ioTailwind CSSVercel

Tech Job Board

AI-powered tech job board with intelligent resume matching and personalized insights. Aggregates remote US tech jobs, stores them in PostgreSQL with pre-computed embeddings for fast matching (~15-20 seconds). Resume upload enables personalized job rankings using a hybrid ML algorithm: 40% skills overlap, 35% semantic similarity (Sentence Transformers), and 25% title alignment. Top 5 matches (80%+ scores) receive AI-generated explanations via GPT-4o-mini highlighting aligned skills and growth opportunities.

FastAPINext.jsPostgreSQLSentence TransformersOpenAI GPT-4o-miniLangChainscikit-learnPythonTypeScriptTailwind CSSRailwayVercelGitHub Actions

Research Outline Generator

Multi-agent AI research system that autonomously investigates topics and generates comprehensive outlines with detailed summaries. Features LangGraph orchestration of specialized agents: a search agent using Tavily API for web discovery, a scraping agent for content extraction, an analysis agent for information synthesis, and an outline generation agent. Built with sequential workflow architecture for reliable research pipeline execution and structured output generation.

FastAPILangGraphLangChainStreamlitOpenAI GPT-4o-miniTavily APIRailway

Catalyst AI - Business Assistant

AI-powered business intelligence assistant leveraging RAG (Retrieval-Augmented Generation) to analyze business data and generate actionable insights. Features interactive visualizations, conversational AI interface, PDF document retrieval with FAISS vector store, and LangSmith-powered model evaluation. Built with modern LLMOps best practices.

StreamlitLangChainOpenAI GPT-3.5FAISSLangSmithPlotlyPandasPythonRAGRailway

US Constitution Assistant

AI-powered conversational chatbot that answers questions about the US Constitution using RAG (Retrieval-Augmented Generation). Features PDF document processing with automatic text chunking, OpenAI embeddings for vector representation, and MemoryVectorStore for similarity-based retrieval. Includes conversation memory for context-aware follow-up questions and accessible, responsive design.

Next.jsLangChainOpenAI GPT-4o-miniRAGMemoryVectorStoreOpenAI Embeddingspdf-parseTailwindReactJavaScriptVercel

NewsGenie

AI-powered news and information assistant that combines real-time news updates with conversational AI. Features smart query classification, multi-source integration (GNews + Tavily), and LangGraph workflow orchestration for intelligent routing between news requests and general queries.

LangGraphStreamlitOpenAI GPT-4o-miniGNews APITavilyPythonRailway

AI Chat Assistant

AI-powered conversational chatbot leveraging Groq's ultra-fast LLM inference with Llama 4 Scout for near-instantaneous responses. Features conversation memory for context-aware interactions, Serper API integration for real-time web search capabilities, and a clean, responsive interface. Optimized for speed and user experience with streaming responses and minimal latency.

Groq with Llama 4 ScoutSerper APINext.jsTailwindReactTypeScriptVercel

VisionAssist

Real-time object detection web app designed to assist visually impaired users. Detects 80 common objects using YOLOv8n model with audio announcements. Features privacy-first architecture with all processing happening client-side in the browser, achieving 20-30 FPS with ~50ms inference time. Includes smart audio feedback that announces consistently detected objects on change or every 5 seconds.

Next.jsReactTypeScriptTensorFlow.jsYOLOv8nTailwind CSSshadcn/uiWebGLWeb Speech APIVercel

Image Generator

AI-powered image generation tool using OpenAI's DALL-E 2 to create custom posters, banners, and visual content from text prompts. Features an intuitive interface for prompt engineering, real-time image generation, and high-quality output suitable for marketing materials and creative projects. Built with responsive design for seamless creation across devices.

OpenAI DALL-E 2Next.jsTailwindReactTypeScriptVercel