David Kiss

• Hired a staff iOS developer (freelance) and an intermediate iOS developer (part-time).
• Transformed the prototype built by an agency into a production-grade mobile app, and its back-end endpoints were built in AWS.
• Designed, built, and maintained back-end endpoints.
• Led technical implementation of new features, including OCR scanning of shoe box labels in order to extract shoe details, ability to follow other users, public/private profiles, email and push notifications, forums, etc.

An AI Agent that provides personalized recommendations on optimizing the user's LinkedIn Profile. It was created using the free low-code https://agent.ai platform.

Key skills: Claude Sonnet 3.7, agent.ai, LinkedIn API

An AI-powered virtual staging platform. I developed an innovative website called "Stage My Dream Home," which leverages generative AI to transform ordinary room photos into beautifully staged spaces in various styles such as modern, classic, and minimalist.

PROBLEM
The project addressed the challenge for realtors to stage empty rooms and create stunning photos of a property while the previous owner/tenant is still living and the property is still cluttered. Traditional staging and post-processing photos are costly and time-consuming, and a more efficient solution was needed.

MY CONTRIBUTIONS
I led the project from concept to completion, utilizing my front-end, back-end, cloud infrastructure, and AI skills.

OUTCOMES
The platform significantly reduced staging costs and time, offering realtors a seamless user experience for creating digitally staged images in a few clicks.

This is a simple demo on how to use the OpenAI Realtime API with Twilio to create a phone bot that can accept incoming phone calls and use OpenAI to generate responses in real time.

Key technologies: OpenAI Realtime API, Twilio Voice, TypeScript, Node.js, Express.js, Websockets, Ngrok

Appointment booking chatbot integration with Google Calendar

I developed a demo chatbot application that utilizes Python, Langchain, MCP, Pipedream, and the Google Calendar API. The project showcases the integration of an appointment booking chatbot with third-party services, specifically Google Calendar, using the Model Context Protocol (MCP) in Langchain.

PROBLEM
The project addressed the challenge of seamlessly integrating chatbots with external services like Google Calendar to automate appointment scheduling and management.

OUTCOMES
The integration successfully automated appointment scheduling, reduced manual input, and streamlined user interactions with Google Calendar. This enhanced user experience demonstrated the potential for broader applications in various service industries.

Created a demo FastAPI application that fetches hotels by city from Booking.com using Playwright, an open-source browser automation library. The project involved setting up a virtual environment, fetching dependencies, and installing the Chromium browser for Playwright. The application was designed to run on a local server using Uvicorn.

Key skills: Python, FastAPI, Playwright

A demo multi-agent chatbot system built for a fictitious Maple Ridge Mortgages company using Python, LangGraph, LangChain, ChatGPT, RAG, and FAISS.

It showcases the common building blocks of Agentic AI:
• RAG (using FAISS)
• Input and output guardrails
• Human in the loop
• Tools - Gmail for sending notification emails and Google Calendar for managing appointments
• Multiple agents with their own responsibilities, including a supervisor, an FAQ agent, a messaging agent, and a calendar agent
• Tracing using LangSmith

Created a demo FastAPI application that integrates with Vapi.ai to make outbound calls to patients and schedule their appointments using a Google Calendar MCP server. This project involved setting up a seamless interaction between AI models and external data sources using the Model Context Protocol (MCP).

Key skills: Python, FastAPI, Vapi.ai, ChatGPT, MCP, Google Calendar

Built this sample project that involved transforming a shopping list application using FastAPI into a Model Context Protocol (MCP) tool for a LangGraph chatbot. The project aimed to enable seamless interaction between AI models and the shopping list application, allowing users to manage their shopping lists via chat.

Key skills: Python, FastAPI, LangChain, LangGraph, ChatGPT, MCP

A sample LangChain based chatbot where users can upload PDF files and ask questions about the PDF's content, including tables.

It uses LangChain's PyMuPDF4LLMLoader to extract text from a PDF file and enriches it with tables extracted using the img2table library.

A sample LLM is used as a judge evaluator to test an AI Agent that can answer questions from an FAQ using LangChain, LangGraph, LangSmith, ChatGPT, OpenEvals, and Pytest. This project aims to show quantitative ways to measure the performance of LLM applications.

This is a WhatsApp customer service and lead generation chatbot for eralpcaner.org that answers users' questions using RAG in English and Turkish and uses call-to-action questions to guide users to the next step in the conversation.

Key skills: n8n, ChatGPT, Pinecone, RAG, tools