Jun Okaniwa
I design and build data systems and retrieval applications that help engineers find technical facts and check where each one came from.
I work as a Software System Designer in automotive software development. Alongside that role, I designed and built the core of an internal system that cross-checks requirements against implementation data: the data model, fixed SQL queries, a reusable Python-to-PostgreSQL pipeline and a FastAPI backend, deployed to Azure. With human review, its database queries have been applied to four related projects.
In my own portfolio project, Evidence-First RAG, a language model helps find the right entity, the person makes the choice, and every fact comes from registered SQL over PostgreSQL.contract It runs on synthetic data and is deployed to Azure App Service.deploy records
Next, I want my main work to be designing, building and improving applied AI and data platform systems.
The flagship project
Evidence-First RAG is a retrieval tool for engineers who check message, signal and mapping definitions by opening several structured files and comparing them by hand, work that is slow to repeat and hard for another engineer to check.
It answers scoped questions about messages, signals and mappings, asks which version is meant instead of guessing, and says plainly when it cannot answer.contract Each answer shows where its facts came from.conformance records It treats the same name in two versions of the data as two different entries.synthetic data