Ever spent hours searching for the perfect tech conference or hackathon? What if an AI agent could do that work for you, delivering personalized event recommendations with detailed analysis in minutes? Here's how I built Event Hunter AI - a sophisticated multi-agent system that transforms how developers find industry events.
The Problem: Event Discovery is Broken ๐คฏ
As developers, we've all been there. You want to attend a conference, find a hackathon, or discover networking events in your field, but:
- Google searches return outdated results from 2019
- Event listing sites are cluttered and hard to filter
- CFP deadlines are buried in dense text
- Sponsorship opportunities are unclear or missing
- You waste hours manually checking dozens of websites
What if we could solve this with AI? Not just any AI - but a multi-agent system that mimics how a human research assistant would tackle this task.
The Solution: Event Hunter AI โก
Event Hunter AI is a sophisticated system that combines:
- ๐ Multi-agent architecture with specialized search and analysis agents
- ๐ Web scraping capabilities via BrightData's Model Context Protocol (MCP)
- ๐ง GPT-4.1-mini for intelligent processing and decision-making
- โก Real-time streaming via WebSocket for live updates
- ๐จ Modern React frontend with form-based query building
Watch the Demo
Why Multi-Agent Architecture? ๐ค
Instead of building a monolithic AI system, I chose a multi-agent approach inspired by how human teams work:
1. Event Search Agent ๐ต๏ธโโ๏ธ
- Role: Find events across the web
- Tools: Search engine access via MCP
- Output: List of potential events with URLs
2. Event Details Agent ๐
- Role: Extract comprehensive information from event pages
- Tools: Web scraping via MCP
- Output: Structured event data (dates, CFP status, sponsorship)
3. Main Orchestrator ๐ฏ
- Role: Coordinate the other agents and compile final results
- Intelligence: Understands user intent and manages workflow
This separation of concerns makes the system more reliable, maintainable, and allows each agent to excel at its specific task.
Tech Stack Deep Dive ๐ ๏ธ
Backend: Python + FastAPI
The heart of the system uses cutting-edge technologies:
# Multi-agent setup with DeepAgents framework
from deepagents import create_deep_agent
from langchain.chat_models import init_chat_model
from langchain_mcp_adapters.client import MultiServerMCPClient
# Initialize MCP tools from BrightData
mcp_client = MultiServerMCPClient({
"brightdata": {
"url": f"https://mcp.brightdata.com/sse?token={brightdata_token}",
"transport": "sse",
}
})
# Create specialized agents
event_search_agent = {
"name": "event-search-agent",
"description": "Finds industry events using search engines",
"prompt": event_search_prompt,
}
agent = create_deep_agent(
tools=mcp_tools,
model=model,
instructions=event_hunter_instructions,
subagents=[event_search_agent, event_details_agent],
)
Key Backend Features:
- FastAPI for modern Python web framework
- WebSocket streaming for real-time updates
- DeepAgents for multi-agent orchestration
- BrightData MCP for web scraping at scale
- Azure OpenAI integration
Frontend: React + TypeScript
A polished user experience with modern web technologies:
// Real-time WebSocket connection
const connectWebSocket = () => {
wsRef.current = new WebSocket('ws://localhost:8000/ws/query')
wsRef.current.onmessage = (event) => {
const data = JSON.parse(event.data)
if (data.type === 'stream') {
setResult(data.content) // Live updates!
}
}
}
// Form-based query building
const query = `Find ${formData.vertical} events in ${formData.location}
${dateRangeText}.${formData.companies ? ` Focus on events where these
companies might be involved: ${formData.companies}.` : ''}`
Key Frontend Features:
- React 19 with TypeScript for type safety
- Tailwind CSS for modern styling
- Framer Motion for smooth animations
- WebSocket integration for live streaming
- Responsive design that works on all devices
The Magic: How It Actually Works ๐ช
Here's the step-by-step process when you submit a query:
Step 1: Query Processing ๐
User Input: "Find AI hackathons in San Francisco for Q1 2025"
โ
System parses: Industry=AI, Location=San Francisco, Time=Q1 2025
Step 2: Search Agent Activation ๐
# Search agent gets to work
search_queries = [
"AI hackathon 2025 San Francisco",
"artificial intelligence hackathon Bay Area",
"machine learning competition San Francisco 2025"
]
# Uses BrightData MCP to search Google, Bing, Yandex
Step 3: Details Agent Analysis ๐
# For each event URL found:
event_data = {
"name": "AI DevFest 2025",
"date": "March 15-17, 2025",
"location": "San Francisco, CA",
"cfp_status": "Open until Feb 1st",
"sponsorship": "Available - contact organizers"
}
Step 4: Real-time Streaming โก
The frontend receives live updates as each agent completes its work:
- "๐ Searching for AI hackathons..."
- "๐ Analyzing event details..."
- "โ Found 12 relevant events!"
Key Features That Make It Special โจ
1. Intelligent Date Awareness ๐
The system knows today's date and filters out past events automatically:
current_date = datetime.now().strftime("%B %d, %Y")
event_search_prompt = f"""
IMPORTANT: Today's date is {current_date}.
When searching for events, always consider this current date
and look for upcoming events, not past ones.
"""
2. Comprehensive Event Analysis ๐ฌ
For each event discovered, you get:
- Event Name and official website
- Exact dates and times
- Location details (city, venue, online/hybrid)
- CFP status - are submissions still open?
- Sponsorship opportunities - can your company get involved?
3. Form-Based Query Building ๐
Instead of trying to craft the perfect prompt, users fill out a simple form:
- Location: San Francisco, Europe, Online
- Date Range: Calendar picker for precise timeframes
- Industry: Dropdown with 12+ verticals
- Companies: Optional company focus
- Additional Requirements: Free-text for special needs
4. Real-Time Processing Feedback ๐ก
Users see exactly what's happening:
๐ Searching for blockchain conferences in Europe...
๐ Found 15 potential events, analyzing details...
โ
Extracted information for 12 high-quality events!
Technical Challenges & Solutions ๐งฉ
Challenge 1: Rate Limiting & Reliability
Problem: Web scraping can be unreliable and rate-limited.
Solution: BrightData MCP provides enterprise-grade scraping with automatic retry logic and proxy rotation.
Challenge 2: Information Extraction Accuracy
Problem: Event websites have inconsistent layouts and data formats.
Solution: Specialized prompt engineering for the Details Agent with clear extraction templates:
event_details_prompt = """
Extract the following information for each event:
- Event Name
- Date(s)
- Location (city, venue if available)
- Main statement/description
- Whether CFP (Call for Papers) is open
- Whether sponsorship opportunities are available
Format as structured markdown for consistency.
"""
Challenge 3: Real-Time User Experience
Problem: AI processing takes time, users need feedback.
Solution: WebSocket streaming with filtered content to show only meaningful updates.
Challenge 4: Scalability
Problem: Each query requires multiple web requests and AI processing.
Solution: Async Python with FastAPI handles multiple concurrent requests efficiently.
Results & Performance ๐
After building and testing Event Hunter AI:
Speed: Average query completion in 60-90 seconds
Accuracy: 85%+ relevance for discovered events
Coverage: Finds 10-15 high-quality events per query
User Experience: Real-time updates keep users engaged
Sample Output Quality
For a query like "Find fintech conferences in London for 2025", the system returns:
**Event Name**: FinTech World 2025
**Link**: https://fintechworld.com
**Date**: April 22-24, 2025
**Location**: ExCeL London, UK
**Main Statement**: Europe's largest fintech conference bringing together 5,000+ professionals
**Open CFP**: Yes - submissions close March 1st, 2025
**Open Sponsorship**: Yes - platinum packages available
Lessons Learned ๐
1. Multi-Agent Architecture is Powerful
Breaking down the problem into specialized agents made the system more reliable and easier to debug. Each agent has a clear responsibility and can be optimized independently.
2. Real-Time Feedback is Crucial
Users will wait 90 seconds for good results, but only if they see progress. The streaming WebSocket updates transformed the user experience.
3. Form-Based Input > Prompt Engineering
Instead of expecting users to craft perfect prompts, a structured form with dropdowns and date pickers made the system accessible to everyone.
4. Modern Tooling Matters
- BrightData MCP: Made web scraping reliable and scalable
- DeepAgents: Simplified multi-agent orchestration
- FastAPI: Handled WebSocket streaming beautifully
- React + TypeScript: Enabled rapid frontend development
Future Enhancements ๐
The current system is just the beginning. Here's what's on the roadmap:
1. Personalization Engine ๐ฏ
- User profiles with preferred industries and locations
- ML-powered recommendations based on past searches
- Calendar integration to avoid scheduling conflicts
2. Advanced Filtering ๐
- Price range filtering (free vs. paid events)
- Event size preferences (intimate vs. large conferences)
- Format preferences (online, hybrid, in-person)
3. Notification System ๐ข
- Email alerts for new events matching user criteria
- CFP deadline reminders
- Early bird pricing notifications
4. Community Features ๐ฅ
- Event reviews from past attendees
- Networking opportunities - see who else is going
- Group discounts for multiple registrations
Getting Started ๐โโ๏ธ
Want to try Event Hunter AI or contribute? Here's how:
Prerequisites
- BrightData account for MCP web scraping
- Azure OpenAI access for GPT-4.1-mini
- Python 3.8+ and Node.js 18+
Quick Setup
# Clone and setup backend
git clone https://github.com/brightdata/event-hunter
cd event-hunter
pip install -r requirements.txt
cp .env.example .env # Add your API keys
# Setup frontend
cd event-hunter
npm install
npm run dev
# Start the system
python main.py # Backend on :8000
# Frontend auto-opens on :5173
Environment Configuration
# .env file
BRIGHTDATA_API_TOKEN=your_token_here
OPENAI_API_KEY=your_azure_openai_key
OPENAI_BASE_URL=https://your-resource.openai.azure.com/openai/v1/
Impact & Use Cases ๐ก
Event Hunter AI isn't just a cool tech demo - it solves real problems:
For Individual Developers ๐จโ๐ป
- Save 5+ hours per event search
- Discover niche events you'd never find manually
- Track CFP deadlines automatically
- Find sponsorship opportunities for side projects
For Companies ๐ข
- Identify conference speaking opportunities
- Find sponsorship targets aligned with their audience
- Track competitor presence at industry events
- Plan marketing calendars around key events
For Event Organizers ๐
- Competitive analysis of similar events
- Market research for new event concepts
- Partnership identification with complementary events
The Bigger Picture ๐
Event Hunter AI represents a new class of AI applications:
1. Multi-Modal Intelligence ๐ง
Combining web search, content analysis, and structured data extraction into a cohesive workflow that mimics human research patterns.
2. Real-Time AI Experiences โก
Moving beyond static chatbots to dynamic, streaming AI that provides progress updates and builds trust through transparency.
3. Domain-Specific AI Agents ๐ฏ
Rather than general-purpose AI, specialized agents that deeply understand specific problem domains (in this case, event discovery).
4. Human-AI Collaboration ๐ค
The system augments human capabilities rather than replacing them - users still make the final decisions about which events to attend.
Conclusion: The Future of AI-Powered Discovery ๐ฎ
Building Event Hunter AI taught me that the future of AI isn't about replacing human judgment - it's about augmenting human capabilities with intelligent automation.
The multi-agent architecture proved that complex problems require specialized solutions. By breaking down event discovery into search and analysis phases, each agent could excel at its specific task while working together toward a common goal.
Most importantly, this project shows how modern AI tooling - from BrightData's MCP to DeepAgents to real-time streaming - makes it possible for individual developers to build sophisticated systems that were previously only feasible for large teams.
What's Next?
The techniques used in Event Hunter AI can be applied to many other discovery problems:
- Job search with salary analysis and company culture insights
- Product research with competitive analysis and pricing trends
- Academic paper discovery with relevance scoring and citation analysis
- Travel planning with weather, events, and local recommendations
Try It Yourself! ๐ฎ
Event Hunter AI is open source and ready to use. Whether you're looking for your next conference or want to learn about multi-agent AI systems, give it a try:
๐ GitHub Repository: https://github.com/brightdata/event-hunter
๐ Live Demo: Coming soon!
๐ Documentation: Detailed setup guide in the README
Have questions about the implementation or want to contribute? Drop a comment below or open an issue on GitHub!
What kind of AI-powered discovery tools would you build? Share your ideas in the comments! ๐
Tags
#ai
#python
#fastapi
#react
#typescript
#multiagent
#websockets
#eventdiscovery
#webscraping
#machinelearning
#automation
#opensource
Top comments (0)