Open to opportunities

Liuyang Wang

@liuyangwang

Senior software engineer building scalable GenAI systems that turn shopping conversations into measurable customer value.

United States

Message

What I'm looking for

I’m looking to lead end-to-end GenAI and platform engineering—building scalable LLM/RAG systems, full-stack user experiences, and reliable cloud/data foundations in teams that value quality and measurable customer impact.

I’m a Senior Software Engineer (L6) at Amazon, focused on end-to-end Generative AI engineering across full-stack platforms, cloud infrastructure, and petabyte-scale data systems. On Rufus, I design and deliver core components of Amazon’s flagship GenAI shopping assistant on AWS Bedrock—driving conversational discovery and agentic auto-buy capabilities for 250M+ customers with projected $10B+ in incremental annual sales.

I lead complex systems from React/Next.js experiences through distributed LLM inference clusters, Parallel RAG pipelines, and enterprise data lakehouses. I’ve reduced inference cost by 28% with speculative decoding, achieved sub-1-second P99 time-to-first-token with Parallel RAG, scaled production inference across 80,000+ Trainium/Inferentia accelerators with zero degradation, and built LLM evaluation and safe deployment workflows that support weekly releases while sustaining 210% YoY interaction growth.

Experience

Work history, roles, and key accomplishments

Current

Senior Software Engineer

Current

Amazon

Aug 2022 - Present (3 years 9 months)

Serves as Senior SDE (L6) on Amazon’s Rufus GenAI shopping assistant, impacting 250M+ customers and projected $10B+ in incremental annual sales. Led end-to-end delivery with Parallel RAG and Bedrock Agents, including speculative-decoding model routing that reduced inference cost by 28% and multi-region scaling across 80,000+ Trainium/Inferentia for 3M tokens/min at Prime Day peaks.

Amazon Bedrock Agents Speculative Decoding VLLM Triton Inference Server React Next.js gRPC GraphQL

Senior Software Engineer

Compass

Jan 2019 - Aug 2022 (3 years 7 months)

Architected AI/ML capabilities for the Compass Platform, migrating 500TB to a Databricks Lakehouse (Delta Lake/Unity Catalog/MLflow). Built recommendation and predictive pricing models achieving sub-5% MAPE and designed AI search ranking with OpenSearch and Kafka exactly-once semantics that reduced index build time from 8 hours to 1 hour.

Delta Lake MLFlow Unity Catalog Apache Spark Kafka OpenSearch Python Great Expectations React

Software Engineer

Google

Nov 2016 - Mar 2018 (1 year 4 months)

Developed TensorFlow learning-to-rank models improving NDCG by 8% and built BERT-based NLP pipelines that reduced query reformulation rate by 15%. Built Apache Beam (Dataflow) pipelines for 10TB+ daily search logs and delivered Java backend services for search ranking signals at 500K+ RPS with P99 < 50ms.