AI in Action

Dynamic AI Companion Powered by Hierarchical Memory

What if an AI could remember you—not just your last question, but your preferences, past conversations, and unique personality? In this episode, host Robert is joined by AI Engineer Terresa Pan to discuss EchoStar, a groundbreaking hierarchical memory system for AI. Terresa explains how different "memory banks" (episodic, semantic, and procedural) work together to create an AI with a genuine, long-term memory, allowing it to truly listen, adapt, and build an evolving relationship with its user.

Insight Canvas & LinkedIn Post Generator Powered by Google ADK

Ever wondered how AI can seamlessly juggle text, images, and data analysis all at once? On this episode of AI in Action, we're unlocking the secrets behind Google's Agent Development Kit with expert Teresa Pan. She's here to discuss how ADK's unique approach to 'artifacts' is transforming the way developers build sophisticated AI agents. We'll explore her fascinating applications, from an AI-powered research assistant to a creative LinkedIn content generator. Stay tuned for a look under the hood of next-generation AI!

Grand Code Agent with Smolagent Framework

AI is evolving rapidly, moving beyond information processing to tackle complex tasks. Today, we explore the "Grand Code Agent," an AI by Teresa Pan that showcases this shift. Forget rigid commands; this agent uses Python code to interact with tools, retrieve financial data, visualize information from PDFs and images, and even generate memes. We'll delve into how this "code-first" approach, powered by the SmolAgents framework, offers greater flexibility, efficiency, and even self-debugging capabilities, changing how AI gets things done. Join us for a behind-the-scenes look at this fascinating technology.

AI Multi-Agent Debate Simulation with Great Minds

This episode discusses an AI Multi-Agent Simulation Project, exemplified by an "AI Debate Simulator." It utilizes Large Language Models (LLMs), grounded by Retrieval-Augmented Generation (RAG) with a Pinecone vector database, to simulate discussions among AI personas representing various individuals. The system allows users to configure simulations with different participants and topics, observe the multi-round interactions, review the knowledge sources used, and pose follow-up questions. Built using Langchain, LangGraph, and Streamlit, the application features a user interface for setup and review, and leverages backend services like Google Generative AI for language processing and Pinecone for knowledge storage.

AI Powered SmartMatch Staffing Platform

The podcast introduces an AI-powered staffing platform called Interactive SmartMatch made by Terresa Pan, an AI engineer and consultant. The platform aims to improve internal efficiency by using AI to optimally match consultants to projects based on skills, availability, and project needs. Key features include automatic extraction of project requirements, a human-in-the-loop verification process, real-time consultant matching with a scoring system, and the use of technologies like Google Gemini, LangChain, and Supabase for its functionality. For more information, visit Terresa Pan's AI Agent Garden [https://ai-agents-garden.lovable.app/]