Kevin Cheng

Engineered production GenAI capabilities for customer workflows by integrating LLM inference, semantic retrieval, and enterprise data. Built AI/ML services in Python and Java supporting large-scale inference with sub-500 ms P95 latency, including distributed serving and ranking. Developed embedding-based search over 1M+ documents to improve relevance and reduce query latency. Created distributed data pipelines using Apache Beam, Dataflow, BigQuery, and Airflow (100+ GB daily), and improved pipeline delays. Built ML-ready feature generation pipelines for standardized model training data, automated multi-step compensation workflows, and architected real-time integrations among enterprise systems. Developed quota and territory applications, designed compensation rule engines for regional business logic, and mentored engineers through reviews and workshops.

Built backend services for cashback and coupon workflows using Python, Flask, Django, and Go, supporting offer ingestion, eligibility, and redemption for high-traffic mobile users. Developed data pipelines for partner catalogs, clickstream, and transaction events to improve offer freshness and reconciliation. Designed PostgreSQL and Redis-backed coupon eligibility and transaction processing services. Automated partner environment provisioning with Terraform and AWS to reduce onboarding time, and refactored transaction services with audit logging and encryption to improve reliability and privacy.

Developed Spring Boot transaction-processing services for batch payment workflows, improving throughput. Built Django, Celery, and Redis ETL pipelines for financial data aggregation, processing 500K+ transactions daily. Optimized PostgreSQL query performance and indexing strategies for nightly reconciliation workloads, reducing processing time. Led migration of legacy SOAP services to REST APIs to reduce integration complexity and maintenance overhead for external partners.