Risk workflow/rule language and execution engine

• Architected a distributed graph database with a SQL like query language. We designed from the storage layer to the query engine from scratch.

• Architected a new payment processing engine, which is a highly reliable, highly scalable fault tolerant system. It is a workflow like system built on top of Waltz distributed log. It stores execution states in Waltz for durability and failover. The entire system is designed for consistency guarantee in the active-active multi-region deployment in mind.

Architected Waltz, a distributed write-ahead log, which is built for multi-region transaction consistency.

blog: https://wecode.wepay.com/posts/waltz-a-distributed-write-ahead-log

code: https://github.com/wepay/waltz

Designed and implemented Kafka Streams. Kafka Streams is a stream data processing library. It is a part of Apache Kafka.

Focused on the design and implementation of the following areas.

• KStream/KTable DSL design

• Time-based Flow control

• Topology analysis

• Task distribution/balancing

• integration with Kafka's consumer group protocol

• fast fault recovery using standby tasks

Architected Kifi search engine, which uniquely combine personalization and social aspect. A user creates keeps (which are like bookmarks) of web pages. The search engine provides full-text search of keeps. Page titles can be personalized, and pages can be tagged. Personalized title and tags are also searchable by the user, but not by other users. A keep can be public or private. If public, the user's social connections can discover the page in search results, and this way the user can help friends finding useful/interesting information. If a keep are private, the fact that the user kept the page is hidden from other users, so the users privacy is protected.

The engine deals with content search, personalized metadata search (personalized titles, tags, and more!) and social graph processing at query time in distributed manner.