About

I’m Junshu Sha, an AI application engineer focused on agents, RAG, and data systems. My background in IT audit and data analysis at KPMG Shanghai shaped how I think about permissions, evidence, control boundaries, and traceability.

How I build AI applications

I build enterprise knowledge platforms, audit automation, and generative-AI workflows that place model capabilities inside systems that can be tested, deployed, and traced back to sources. Rather than presenting a single successful model response, I care about how data enters the system, how answers return to evidence, where authorization takes effect, and how failures can be found and reproduced.

Current focus

• My Company Brain: A multi-knowledge-base RAG and Agent platform for documents, relational knowledge, knowledge pages, and an assistant.
• AI Mathematical Animation Workbench: Turns teaching requests into structured plans, then generates, repairs, and validates Manim animations in a constrained execution environment.
• ITA-Maskit: A local masking tool for audit materials, with deterministic pseudonymization, decoupled rules, and million-row processing.