Shiying Li

Designed and scaled a HIPAA-compliant healthcare data lakehouse on S3, Lake Formation, Glue, Apache Iceberg, and Redshift to centralize EHR, clinical, claims, scheduling, and operational data for a unified patient view and population-health analytics. Built serverless, event-driven ingestion pipelines with Lambda, API Gateway, FHIR R4, DMS (CDC), and EventBridge to process API and CDC workloads and improve data freshness with high availability. Developed transformation layers using dbt, Redshift, SQL, PySpark, and GitHub Actions to create reusable star-schema and patient-centric data marts, reducing report generation time from hours to minutes. Implemented near-real-time pipelines with MSK/Kafka, Kinesis, Lambda, and EventBridge so downstream consumers could access encounters, results, claims, and patient events within minutes. Deployed automated data quality and anomaly detection with Great Expectations, CloudWatch, Glue, and Python to detect delays, duplicates, schema drift, and reco

Designed and scaled an AWS healthcare data platform using HealthLake, S3, Glue, Lake Formation, Apache Iceberg, and Athena to consolidate EHR, claims, laboratory, pharmacy, and clinical data into standardized longitudinal patient datasets. Architected regional data isolation and privacy controls with Lake Formation, IAM, KMS, CloudTrail, and S3 encryption to support least-privilege access, auditability, data residency, retention, and controlled processing for sensitive healthcare data. Built high-throughput ingestion pipelines (batch/API/CDC) using Airbyte, Lambda, API Gateway, DMS, EventBridge, and S3, improving freshness and traceability. Developed a FHIR R4 interoperability layer mapping proprietary schemas into standardized resources. Engineered distributed ETL with Spark/Glue for identity resolution, terminology mapping, deduplication, incremental processing, and late-arriving records. Implemented privacy-aware workflows (tokenization/pseudonymization/encryption) and built analyti

Extended Notion’s core data platform to support structured databases so underlying records could power tables, Kanban boards, calendar views, lists, galleries, and page-based views without duplicating source data. Designed flexible metadata and property models using PostgreSQL, SQL, JSON, UUIDs, and relational references to enable dynamic user-defined schemas. Built reusable query and transformation logic for consistent filtering, sorting, grouping, aggregation, and relational-property resolution across multiple views. Engineered relationship models with referential integrity for connecting records and supporting increasingly complex workspace structures. Optimized PostgreSQL indexes and query execution for large dataset views to improve interactive performance, and delivered schema migration/backfill workflows to introduce new properties without disrupting existing content.