Inspiration
In today’s fast-paced digital economy, businesses struggle to stay ahead of competitors due to the overwhelming amount of fragmented data across markets. We were inspired to solve this by designing a system where autonomous agents collaborate to monitor, analyze, and report on market activity — giving businesses real-time, actionable insights without manual effort.
What It Does
Market Pulse is a multi-agent system built with the Agent Development Kit (ADK) that:
- Collects real-time data from online sources (e.g., competitor pricing, product reviews, news)
- Analyzes trends, sentiment, and anomalies using ML/NLP
- Queries historical data via BigQuery to identify patterns
- Generates and delivers automated reports
- Sends alerts when business-critical thresholds are met
Each agent specializes in a specific task, collaborating to produce timely market intelligence.
How We Built It
We used the Python version of ADK to design and implement five core agents:
- ScraperAgent: Pulls competitor/product data from public APIs
- BigQueryAgent: Retrieves historical market and sales data from Google BigQuery
- AnalyzerAgent: Applies NLP and sentiment models to identify trends and anomalies
- ReporterAgent: Compiles daily insights into visualized reports
- NotifierAgent: Sends real-time alerts via Slack or email
Agents communicate through Google Pub/Sub, and outputs are stored in Google Cloud Storage or delivered directly to users. All agents are containerized with Docker and can be deployed via Google Cloud Run.
Challenges We Ran Into
- Learning how to structure and orchestrate agents using ADK
- Ensuring reliable communication through Pub/Sub with message deduplication and retries
- Securely connecting and authenticating with multiple Google Cloud services
- Designing a modular report generator that could handle dynamic insights
- Tuning NLP models to deliver accurate and relevant sentiment analysis
Accomplishments That We're Proud Of
- Deployed a fully operational multi-agent system using ADK
- Enabled real-time analysis and alerting from live market data
- Created a reusable system for competitive intelligence automation
- Demonstrated robust agent collaboration with minimal supervision
What We Learned
- Best practices for building autonomous, modular agents
- How to utilize Google Cloud tools (BigQuery, Pub/Sub, Storage) with ADK
- Techniques for fault-tolerant messaging in distributed systems
- How to bridge machine learning outputs with business-oriented decision-making
What's Next for Market Pulse
- Develop a real-time web dashboard for insight visualization
- Allow customizable rules and scheduling for industry-specific alerts
- Expand data sources to include social media, competitor ads, and job listings
- Integrate Vertex AI for predictive modeling and deeper insights
- Open-source the framework to enable community-driven agent contributions
Log in or sign up for Devpost to join the conversation.