December 1, 2025
Overview
By the numbers
- 210,000+ users
- 11.5M+ referenced creators
- 30M+ indexed posts
- Massive data: creators, campaigns, performance across all social networks
- Real-time: messaging, analytics, dynamic widgets
- Artificial intelligence: LLMs, semantic search, autonomous agents
- Automation: outreach, contracts, reporting
Frontend contributions (Next.js / React)
Architecture & Patterns
- Implementation of a large-scale modular feature-based architecture
- Definition of standard patterns:
.api→.hook→.type→.socket - Strong focus on performance (lazy loading, virtualization, streaming)
Advanced state management
- React Query: server cache, smart invalidation, optimistic updates
- Redux Toolkit: persistent global state for cross-feature data
Real-time
- Implementation of Socket.io and SSE for live updates
- Token-by-token streaming for a "live typing"-style UX on AI responses
- Drag & drop widgets with real-time synchronization
Complex features
- A global cross-feature search system
- Configurable drag & drop widgets
- Messaging interfaces with AI streaming
Backend contributions (Node.js / Express)
Critical business modules
- Campaigns: complete management of influencer campaigns
- Discovery: creator search and filtering
- Messaging: communication with influencers
- Reports: generation of analytics reports
Architecture
- Service / Controller / Repository pattern
- Integration of databases suited to different use cases (SQL, NoSQL, vector DB, cache)
Performance
- Multi-level caching
- Optimization of real-time connections (SSE, WebSockets)
Krown, the B2C product (tech lead)
- Monorepo architecture: SwiftUI app, NestJS API, Next.js web and shared contracts generated in Swift, TypeScript and Kotlin
- Koach, an AI coach powered by an MCP server
- App Store submission, StoreKit 2 billing and release process
AI & Agent systems
Multi-agent (LangChain / LangGraph)
- Development of orchestrated multi-agent systems
- Specialized agents: channel selection, contextual analysis, conversational memory
- Intelligent orchestration of AI workflows
AI streaming
- SSE streaming for real-time AI responses
- Token-by-token display on the frontend
- Handling of interruptions and context
Semantic search
- Qdrant integration with embeddings
- Creator search by semantic similarity
- Continuous improvement of result relevance
Automated generation
- Personalized messages for outreach
- Influencer contracts
- Intelligent creator recommendations
AI observability
- Tracing of AI requests
- Performance and quality metrics
- Feedback system for continuous improvement
Tech stack
Frontend
- Framework: Next.js, React, TypeScript
- State: React Query, Redux Toolkit
- UI: Radix UI, custom components
- Real-time: Socket.io client, SSE
Backend
- Runtime: Node.js, Express
- Databases: SQL, NoSQL, Vector DB, Cache
- Architecture: modular services
Mobile
- iOS: Swift (native application)
AI
- Frameworks: LangChain, LangGraph
- Capabilities: Embeddings, semantic search, agents
DevOps
- Git, Docker, CI/CD
Challenges and lessons learned
- Working on a substantial codebase with high quality standards
- Designing scalable frontend architectures for complex features
- Integrating LLMs in production under latency and cost constraints
- Managing real-time at scale (WebSocket, SSE, streaming)
- Collaborating effectively with product, design, and data teams
Impact
- Significant improvement of the user experience thanks to AI streaming
- Reduced creator search time thanks to semantic search
- Frontend architecture that became an internal reference for new features
What this experience brings
- Work on a complex SaaS product in production
- Master large-scale frontend & backend architectures
- Integrate LLMs and agentic systems into a real product
- Develop an approach oriented toward performance, UX, and maintainability
- Grow on real-time, streaming, and applied AI challenges
