AI Agents Website builder
AI Agents Website builder
Agentic Website Builder is a multi-agent system that automatically generates a complete, deployable full-stack website from a simple user description. The system is built using LangGraph to orchestrate multiple specialized AI agents, each responsible for a different stage of the development process. The workflow begins with a Coordinator agent that manages the pipeline and passes the project description to a Product Manager (PM) agent that defines requirements and structure. A Designer agent then selects an appropriate design pack consisting of themes such as modern-minimal, bold-tech, editorial, or warm-landing, along with layout systems like minimal-stack, split-showcase, bento-grid, editorial-columns, timeline-flow, or sidebar-pro to create varied and visually distinct UI structures. The Frontend agent generates the user interface using a token-based design configuration (design_tokens.json) and a section schema (section_schema.json), which dynamically renders the final website structure in index.html. After the frontend is created, a Quality agent performs heuristic checks and automatically applies safe fixes while storing results in a quality report. The Backend agent then builds a FastAPI service that supports features such as form submissions, including a POST /api/booking endpoint that stores requests in a persistent JSON data file. A DevOps agent prepares the project for deployment, while Validator and QA agents ensure the generated site meets structural and functional standards. Throughout the process, the system provides real-time progress updates in the terminal, showing the activity of each agent (e.g., Coordinator, Planner, Frontend, Backend, DevOps, Validator), giving visibility into the entire automated website creation pipeline. GitHub link
LLM Guardrails
LLM Output Guardrails Service is a FastAPI-based security layer designed to inspect and control LLM-generated responses before they are delivered to end users. The service acts as a protective middleware between the language model and the user interface, ensuring that sensitive, harmful, or invalid content is filtered or corrected. It performs multiple validation checks on model outputs, including moderation for harmful content such as toxicity, hate speech, violence, self-harm, or sexual material, which are automatically blocked. The system also detects personally identifiable information (PII) such as emails, phone numbers, credit card numbers, and Social Security numbers, masking them with placeholders like [REDACTED_*] to prevent data leakage. In addition, the service scans for secrets and credential-like patterns, including API keys (OpenAI, AWS, GitHub), database connection strings, and other sensitive tokens, blocking responses that expose such information. For applications that require structured responses, the service can validate LLM output against a user-provided JSON schema, attempting automatic repairs when possible and blocking responses if they remain invalid. The project also includes a simple RAG-style demonstration interface where users can upload documents and test how the guardrails filter or modify responses, illustrating how the system protects AI applications by enforcing safety, privacy, and output reliability.
AI Academy
This project is an interactive web game that teaches AI concepts to children aged 7–12 through fun mini-games. Built with React, TypeScript, Three.js, and TensorFlow.js, it creates an engaging learning environment where kids explore ideas like pattern recognition, classification, data labeling, and decision trees. The platform features a 3D academy-style hub, smooth animations, and hands-on activities such as training a simple model in the browser and solving AI-themed challenges. A Chatbot Buddy provides recaps and quizzes, helping reinforce concepts while demonstrating how modern AI-powered applications work.
Biz Insights
BizInsights is an AI-powered web application designed to help users explore and analyze business datasets quickly and intuitively. Users can upload CSV or JSON files directly through the interface, after which the system automatically profiles the dataset by identifying row and column counts, sample records, and detailed column statistics. The platform detects common data types such as numeric, categorical, and datetime fields and provides useful insights including null counts, distinct values, top categories, and numeric ranges. Using AI, BizInsights generates business insights, recommends key performance indicators (KPIs) with formulas and explanations, and suggests appropriate dashboard visualizations. The application can also render chart recommendations such as bar, line, pie, scatter, area, and histogram charts to help users visually understand their data. Additionally, users can ask natural-language questions about the dataset, and the system returns both a text-based answer and a structured result table, sometimes recommending a chart when visualization would improve understanding. The platform follows a simple workflow where users upload a dataset, review automated data profiling, explore AI-generated insights and KPIs, view suggested dashboards, and query their data in plain English, all within a single web interface powered by a React frontend served by a FastAPI backend.
SightCraft
SightCraft is a full-stack image application built with React, Express, and the OpenAI API that enables users to both analyze and generate images using AI. The platform supports two main workflows: users can upload an image and ask questions about its content using OpenAI’s vision models, or they can generate new images from text prompts using AI image generation models. The application features a modern React frontend with a custom SightCraft-branded interface and an Express-based backend that handles API requests and model interactions. Designed for both development and deployment, the project includes a Heroku-ready setup, making it easy to run locally or deploy as a scalable web application.
Resume Optimizer
Resume Optimizer is a Streamlit-based application that improves and rewrites a Word .docx resume by aligning it with a pasted job description while preserving the original document’s structure and formatting. The user interface is built in web_app.py, while the resume-processing and optimization logic runs in app.py. The tool analyzes the resume and job description to rewrite experience bullets for stronger alignment with the target role, update the professional summary and skills section, and maintain the original Word layout, paragraph structure, and styling during reconstruction. It also performs an ATS-style analysis, identifying missing skills and keyword frequency to improve applicant tracking system compatibility, and automatically generates a tailored cover letter based on the optimized resume and job description.