[ 01 / IDENTITY ]
YINGQIANGYUAN
Building LLM-powered systems that solve real business problems — from agentic applications and retrieval-augmented generation to LLM evaluation infrastructure on AWS.
TALK TO MY AI TWIN
Explore my experience, projects, and discover what makes me different.
[ 02 / FEATURED WORK ]
SHIPPED IN PRODUCTION
Agentic BI for Lending Analytics
An AI agent that turns natural-language questions into multi-step SQL analyses across a seven-table Snowflake schema. Built on AWS Bedrock with Strands Agents and RAG over an S3 vector store, it encodes credit-risk methods like vintage cohort analysis, implied default probability, and borrower similarity scoring as retrievable references the agent can ground every answer in.
Prompt Eval & Adversarial Testing
An LLM-as-Judge evaluation pipeline for an enterprise LLM system. Roughly 200 annotated test cases (25% of them adversarial) cover six use cases and are scored on business correctness plus five security risk dimensions. Wired into GitHub Actions CI with deployment gates and S3-persisted regression metrics, it became the team's first formal prompt QA process before production releases.
Enterprise Data Lake for Fintech
A Bronze, Silver, and Gold lakehouse on AWS for a fintech with 2M cardholders and 800M annual transactions. Real-time and batch ingestion run through Kinesis (2.2M transactions per day), Kafka, and Step Functions, while Delta Lake ACID upserts execute on Lambda using delta-rs and Polars. LakeFormation column-level ACLs and three-tier CloudWatch alerts cut regulatory audit response from three weeks to under twenty-four hours.
[ 03 / GET IN TOUCH ]
LET'S BUILD TOGETHER
I'm a recent MS CS graduate actively looking for new-grad / full-time roles in AI engineering or data engineering. Three internships in, shipping production systems on AWS, and ready to bring that energy to your team.
✦ OPEN TO RELOCATION · BASED IN SANTA CLARA, CA
ASK MY AI TWIN FIRST