Inspiration Climate change is one of the most pressing challenges of our time, yet many people struggle to understand their personal environmental impact or know how to make meaningful changes. We were inspired by the UN Sustainable Development Goal 13: Climate Action and wanted to create a solution that makes carbon footprint tracking both accessible and financially rewarding. The idea of combining AI coaching with real carbon credits that have monetary value came from recognizing that people need both education and incentives to drive lasting behavioral change. What it does Klymate AI is a comprehensive carbon footprint tracking platform that serves as your personal climate coach. Users can:
Track Daily Habits: Log transport choices, diet, energy usage, and lifestyle activities with real-time carbon calculations Get AI-Powered Coaching: Receive personalized recommendations through natural language conversations with an AI coach that remembers your history and provides contextual advice Earn Real Carbon Credits: Get Klymate Credits (KC) for verified carbon reduction activities that can be converted to cash, carbon offset certificates, or donated to climate causes Compete and Engage: Participate in gamification features including badges, streaks, leaderboards, and community challenges Access Deep Analytics: View detailed insights, trends, and progress toward personalized reduction goals
The platform creates a multi-step agentic workflow where the AI coach analyzes user behavior patterns, searches for relevant sustainability tips using vector search, and automatically suggests personalized action plans. How we built it We built Klymate AI using a modern, scalable architecture centered around TiDB Serverless: Backend Architecture:
FastAPI for high-performance Python web framework TiDB Serverless as our core database with vector search capabilities for storing user data, habit logs, and enabling semantic search for AI recommendations LangChain for AI orchestration and conversation memory management OpenAI GPT models for intelligent coaching conversations SQLAlchemy for async ORM database operations
Multi-Step Agent Workflow:
Data Ingestion: User habits and activities are logged and stored in TiDB with vector embeddings Intelligent Search: Vector search queries TiDB to find similar patterns and relevant sustainability tips LLM Analysis: AI coach analyzes user data and search results to generate personalized recommendations Automated Actions: System automatically updates user goals, awards credits, and triggers notifications
Integration Points:
Firebase for authentication Redis for caching and session management Carbon market APIs for real-time credit valuations Blockchain-style verification for carbon credit transactions
Challenges we ran into
Vector Search Optimization: Tuning TiDB's vector search for semantic similarity between user habits and sustainability recommendations required extensive experimentation with embedding models and similarity thresholds Real-time Carbon Calculations: Creating accurate, real-time carbon footprint calculations for diverse activities while maintaining fast response times AI Context Management: Ensuring the AI coach maintains conversation context and user history across sessions while providing relevant, non-repetitive advice Credit Verification System: Building a transparent, fraud-resistant system for verifying carbon reduction activities and awarding credits Multi-step Workflow Orchestration: Coordinating the complex flow from habit logging through AI analysis to automated credit awards
Accomplishments that we're proud of
Innovative Carbon Credit System: Successfully implemented a working carbon credit economy where users earn real, convertible credits for verified environmental actions Advanced AI Integration: Created a contextual AI coach that uses TiDB's vector search to provide truly personalized recommendations based on user behavior patterns Scalable Architecture: Built on TiDB Serverless, enabling horizontal scaling as the user base grows Multi-Modal Agent Workflow: Achieved seamless integration of data ingestion, vector search, LLM analysis, and automated actions in a single platform Real-World Impact Potential: Created a system that addresses the gap between climate awareness and action through financial incentives
What we learned
Vector Databases in Practice: Gained deep understanding of how vector search can enhance AI applications, particularly in matching user behaviors with relevant sustainability advice Agentic AI Design: Learned to orchestrate complex, multi-step AI workflows that combine data analysis, search, and automated decision-making Financial Incentive Systems: Discovered how monetary rewards can significantly boost user engagement in sustainability platforms Database Performance at Scale: Experienced TiDB Serverless's capabilities for handling both transactional data and vector operations efficiently Climate Action Technology: Deepened our understanding of carbon accounting, verification systems, and the intersection of technology and environmental impact
What's next for Klymate AI
Corporate Partnerships: Integrate with companies looking to purchase verified carbon credits from our user community Mobile Application: Develop native iOS and Android apps with offline capability and location-based tracking IoT Integration: Connect with smart home devices, fitness trackers, and vehicles for automated habit detection Community Features: Expand social aspects with team challenges, local environmental initiatives, and peer-to-peer carbon credit trading Machine Learning Enhancement: Implement predictive models to forecast user carbon footprints and proactively suggest interventions Global Carbon Market Integration: Partner with established carbon exchanges to provide broader liquidity for user credits AI Coach Evolution: Enhance the AI with specialized knowledge domains (renewable energy, sustainable travel, eco-friendly products) and multi-language support
Log in or sign up for Devpost to join the conversation.