Apache Flink + MCP Demo - Real-time data processing using Gemini AI Agent
In this episode Bhargav Kosaraju, Practice Director Data & Streaming at Synapsewerx, shares his recent demonstration from the Melbourne Kafka meet-up. In this demo, Bhargav walks us through a use case and the detailed solution architecture before he shows us the code he's developed to interact with 3 separate MCP Servers, all of which he'll demo live! Bhargav uses an advanced rules-based prompting technique to provide structure and deep context to a Gemini model which is invoked within a Flink Broadcast State Pattern. In this CRM based use case, the LLM uses a business logic to generate an MCP prompt on the fly and in real time. In this demo, the MCP server has multiple action types available to it before updating the event stream. During the demo, business and technical context is provided via discussion with Malcolm Ingram, CEO Synapsewerx and Carlos Rodriguez Iturria, CTO and Chief Architect at Synapsewerx. 00:00:00 - Start 00:00:13 - Welcome & Introduction 00:02:53 - What is an MCP Server 00:07:10 - Demo: Flink MCP Kafka GenAI LLM 00:07:41 - Solution Architecture 00:16:38 - MCP & LLM clarification 00:18:56 - Business benefit of MCP 00:20:07 - MCP Scalability 00:20:33 - Demo: Business Context & Use Case 00:23:57 - The wonder of in-stream inference 00:25:12 - Demo: Solution design 00:31:42 - Rules based LLM constraints 00:33:06 - Demo: Payloads 00:35:54 - Prompts are the new JSON 00:39:42 - Demo continued code walkthrough 00:40:50 - Invest in Kafka schemas! 00:41:44 - Demo continued walkthrough 00:43:24 - Demo execution 00:49:23 - The vendor rush to MCP! 00:51:13 - Demo & Code walkthrough 00:58:04 - Breadcrumbs 00:58:46 - Handling hallucinations 01:01:16 - Designing for failure 01:05:44 - Pay attention: Integration Engineers 01:08:13 - FLINK: Broadcast state pattern 01:12:58 - Confluent Cloud Free Credits 01:13:41 - Closing Comments 01:17:03 - Outro Follow Bhargav Kosaraju / bhargavkosaraju Follow Carlos Rodriguez Iturria / citurria Follow Malcolm Ingram / malcolmingram SWX Social Website: https://synapsewerx.ai LinkedIn: / synapsewerx Confluent Social Website: https://www.confluent.io/ LinkedIn: / confluent This video was not paid for by sponsors. #MCP #Flink #Kafka #Confluent #RealTimeAI #ContextAware #IntelligentTransformation #TheFutureOfIntegration #ThoughtLeadership #AIAgents #AgenticAI

AI Agent Security: Protecting Against Malicious Skills

FULL Claude Tutorial For Beginners in 2026! (FULL COURSE)

From Query to Kafka; How does Apache Flink actually work? by DANNY CRANMER

RL for Agents Workshop - Deep Dive on Training Agents with RL and Open Source

Why MCP really is a big deal | Model Context Protocol with Tim Berglund

Master Real-time Data Pipelines with Kafka and Flink - 3 hr Course - DataExpert.io Free Boot Camp

Run AI Locally: Foundry Local + Microsoft Agent Framework | Udaiappa Ramachandran

Unlocking the Power of Apache Flink: An Introduction in 4 Acts

Model Context Protocol (MCP) Explained for Beginners: AI Flight Booking Demo!

Full Archon Guide - Build AI Coding Harnesses That Actually Ship (LIVE)

Apache Flink: What it is and how it works.

Stream Processing As You’ve Never Seen Before (Seriously): Apache Flink for Java Developers

Don't learn AI Agents without Learning these Fundamentals

Steffen Hoellinger – Flink Jobs as Agents – Stream Processing for Agentic AI #bbuzz

Anthropic Workshop: Build Agents That Run for Hours — Ash Prabaker & Andrew Wilson

Harnesses in AI: A Deep Dive — Tejas Kumar, IBM

MCP Crash Course: What Python Developers Need to Know

Realtime Streaming with Apache Flink | End to End Data Engineering Project

RAG & MCP Fundamentals – A Hands-On Crash Course

