Projects

I have worked on so many projects over the years. This page lists some recent active projects.

Xia

Xia is a secure Portable Persistent AI Assistant (P2A2) for online work. It runs as a single local application, remembers context across sessions, works with any LLM models, and helps with web research, browser automation, authenticated APIs, and recurring online tasks. Xia does not have ambient access to the host machine and runs task in sandbox.

One of the main innovation of Xia is its human-like memory architecture that separate episodic from semantic memories, and a hippocampus for consolidation of knowledge. Xia’s internal state representation separates persistent goals from transient states, and is robust in handling long horizon tasks. The vision of Xia is to become an OS of intelligence.

Datalevin

This is an open-source database that I have been working on since 2020. It is a new generation of general purpose database that is much more ergonomic than traditional SQL RDBMS, while still has the benefits of relational databases (unlike those NoSQL DBs). Datalevin uses a user friendly favor of Datalog as the query language and is good at deductive reasoning and analytics. My goal is not just to compete with existing databases, but also to lay the foundation for advanced artificial intelligence of the future: Just like memory is the center of human cognition, databases ought to serve the same purpose in artificial intelligence.

I have written some articles about Datalevin.

• Achieving High Throughput and Low Latency through Adaptive Asynchronous Transaction
• Competing for the JOB with a Triplestore
• T-Wand: Beat Lucene in Less Than 600 Lines of Code

The work on Datalevin is ongoing. github

Juji Platform

I drove R&D of Juji Platform, an AI conversational agent platform that is the basis of Juji Studio, Juji IDE and Juji API products of Juji Inc..

Juji is the leading conversational agent platform that allows customers to retain total control while using generative AI, a unique yet critical capability that enables successful enterprise use of generative AI.

I have given talks and written about Juji platform.

• How can AI like ChatGPT be turned into gold in enterprises? in 2023
• Generative AI: Past, Present, and Future – A Practitioner’s Perspective in 2023
• Three Big Ideas Behind the Surprising Power and Ease of Use of Juji Platform in 2020
• A Clojure Fusion of Symbolic and Data Driven AI. in 2018

More details can be found on Juji Documentation site.

Editscript

This is an open-source Clojure library I wrote to diff/patch data structures. It is widely in the Clojure community. I have given a talk on its use: Data Diffing Based Software Architecture Patterns

The work on the library is mostly done, but I am still maintaining it. github