The latest articles on DEV Community by Mrunalini pachpute (@mrunalini_pachpute_bf0b2e). https://dev.to/mrunalini_pachpute_bf0b2e https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3936251%2Ffb3c140a-a022-47fa-a8b2-e42f111f0cda.png https://dev.to/mrunalini_pachpute_bf0b2e en Mrunalini pachpute Sun, 17 May 2026 13:06:16 +0000 https://dev.to/mrunalini_pachpute_bf0b2e/i-built-an-ai-ipl-war-room-using-gemini-multi-agent-reasoning-1af7 https://dev.to/mrunalini_pachpute_bf0b2e/i-built-an-ai-ipl-war-room-using-gemini-multi-agent-reasoning-1af7 <p><strong>What if an IPL captain had an AI tactical war room?</strong><br> Not a chatbot:)</p> <p>A system where multiple AI agents debate:</p> <ul> <li>who bowls next</li> <li>whether Bumrah should bowl now or later</li> <li>when to attack</li> <li>how dew changes strategy</li> </ul> <p>That idea became CaptainCool AI — a multi-agent IPL strategist powered by Google Gemini.</p> <p><strong>🧠 The Core Idea</strong><br> Most sports AI apps generate one generic answer.</p> <p>But real cricket decisions involve:</p> <ul> <li>disagreement</li> <li>tactical tradeoffs</li> <li>risk analysis</li> <li>momentum shifts</li> </ul> <p>So instead of one AI model pretending to do everything, I built:<br> an AI captaincy war room.<br> ⚡ Multi-Agent System</p> <p>CaptainCool AI uses multiple Gemini-powered agents:</p> <p>1.🧠 Strategist<br> Proposes the tactical move.</p> <p>2.📊 Stats Analyst<br> Validates the decision using cricket context and live match state.</p> <p>3.🔥 Devil’s Advocate<br> Challenges risky plans and forces reconsideration.</p> <p>4.👑 Final Decision Engine<br> Combines all debate outcomes into the final captaincy call.</p> <p>5.🎙️ Commentary Agent<br> Turns the reasoning into IPL-style live commentary.</p> <p>The result feels far more human than a normal chatbot.</p> <p><strong>CaptainCool AI operates in two different modes:</strong></p> <h3> 🧪 Manual Tactical Mode </h3> <p>Users can manually simulate IPL scenarios by entering:</p> <ul> <li>score</li> <li>wickets</li> <li>overs</li> <li>pitch conditions</li> <li>dew factor</li> <li>captain style</li> <li>impact player availability</li> </ul> <p>This mode was designed for cinematic tactical simulations and reliable demo scenarios.</p> <h3> 🔴 Live Match Beta Mode </h3> <p>I integrated CricAPI to fetch live cricket matches and auto-fill the tactical dashboard in real time.</p> <p>The system processes:</p> <ul> <li>score</li> <li>wickets</li> <li>overs</li> <li>venue</li> <li>batting side</li> <li>match pressure</li> </ul> <p>…and feeds that directly into the Gemini reasoning pipeline.</p> <p><strong>🎨 Frontend Experience</strong></p> <p>I wanted the app to feel like an IPL broadcast control room.</p> <p>So the UI includes:</p> <ul> <li>dark navy gradients</li> <li>cyan + gold accents</li> <li>glassmorphism cards</li> <li>animated confidence bars</li> <li>cricket-ball loading animations</li> <li>sequential agent debate reveals</li> </ul> <p>Built using:</p> <ul> <li>Node.js</li> <li>Express.js</li> <li>EJS</li> <li>Gemini 2.5 Flash</li> <li>CricAPI</li> </ul> <p><strong>🧩 Biggest Challenges</strong></p> <p>1.Multi-Agent Coordination<br> Making agents genuinely disagree instead of repeating similar answers.</p> <p>2.API Quotas<br> Multi-agent reasoning consumes API calls quickly, so responses had to be optimized carefully.</p> <p>3.Live Match Data<br> Live cricket feeds often return incomplete data, requiring normalization and fallback handling.</p> <p>📈 What I Learned<br> The biggest insight was:<br> AI systems become dramatically more believable when agents disagree instead of instantly agreeing.<br> That tactical conflict made CaptainCool AI feel much more realistic.</p> gemini gdgcloud gdgcloudpune gdg