How Vinted Solves Community Support Requests Effectively
summary
Vinted moves from a Ruby monolith to modular services, handling thousands of member requests daily for a platform with 75 million active users by building a three‑layer funnel: self‑help, machine help, agent help. Tech choices include Golang services, Strapi and Phrase for content, Vespa vector search for semantic FAQ retrieval, Kafka for messaging, Temporal for orchestration and a Python LangChain chatbot with anonymization and prompt‑injection guards. Early rollout shows faster, higher‑quality resolutions and fewer escalations to agents.
AI-written summary · about
chapters (10)
- 0:00Introduction to Vinted
- 3:56Community support at Vinted
- 6:24Support challenges and issues
- 9:13Help experience domain creation
- 10:39Support funnel and automation layers
- 13:07Self-help system architecture
- 15:15Machine help layer and chatbot
- 17:03Chatbot service details
- 19:12Agent help and AI integration
- 20:16Results and future plans
related talks
31mThe rocky road from monolithic to microservices architecture
33mAutomating Resolution of IT Support with Chatbots
42mBeyond Vectors: Evolving GenAI through Transformative Tools and Methods
33mBuilding Recommendation Systems for Global E-commerce Marketplaces
06mINTERVIEW
35mScaling RAG Applications to serve millions of users
26mUnleashing the Potential of Cloud Native Open Source Vector Databases
20mUnderstanding Vespa with a Lucene mindset