[team]

Knowledgebase & Contacts CRM, on a real-time event pipeline

Two products built from scratch and still in production for millions of websites, kept in sync by a Kafka and MongoDB Change Streams pipeline.

tawk.to Knowledge Base article editor with document settings for slug, status, visibility, category, language and author — The Knowledge Base article editor, from the live demo on tawk.to (captured November 2026). The product has evolved since my time on the team (2019–2021).

tawk.to Knowledge Base editor with the slash menu open, listing Header, Paragraph, Image, Code and Video blocks — Typing "/" opens the block menu to insert headings, images, code and video. Also from the live demo.

tawk.to Knowledge Base product page showing a help center article with a search bar — The Knowledge Base product page: a help center on the customer's own domain, in 50+ languages.

tawk.to Contacts (CRM) product page showing a contact record with details and an activity timeline — The Contacts (CRM) product page: every visitor who messages becomes a contact with their full history. The contact shown is tawk.to's sample data.

Context

tawk.to customers use the platform to support their own visitors. They needed a self-service knowledgebase and a way to manage the contacts they talk to. Behind both sits tawk.to's backend of distributed services that need to react to each other's data. All of this was built by a small engineering team of 4–5 people working with the CTO.

Problem

Both features had to be built from nothing and hold up under the traffic of a platform used on millions of websites. Core workflows were tightly coupled, and the services needed to learn about data changes quickly and reliably without polling the primary database.

What I did

Worked on the team that built the Knowledgebase service and the Contacts CRM from scratch, alongside the CTO.
Developed backend features for both services in Node.js and MongoDB.
Worked on the team that designed and built an event-driven service on Apache Kafka, decoupling core workflows and giving services a reliable way to communicate.
Helped build the real-time data pipeline that captures changes with MongoDB Change Streams and publishes them to Kafka for downstream consumers.
Implemented Elasticsearch-powered search and filtering.
Kept 100% test coverage on the services I owned, to keep regressions out of production.
Contributed to architecture discussions within the team.

Impact

Both services are still active in production today.
Customers can search and filter their knowledgebase and contacts quickly.
Data changes reach downstream services with minimal latency, and decoupled workflows let the services evolve independently.