Optimizing vLLM Inference Latency: A Deep Dive into PagedAttention
Exploring how memory fragmentation impacts large model serving and how PagedAttention solves the KV cache bottleneck in production environments.
Read on Medium
About Me
Where Research Meets Engineering
Academic Track
I am a Master's student in Artificial Intelligence at the University of Pennsylvania. My research focuses on Large Language Models (LLMs) and Brain-Computer Interface (BCI), aiming to improve LLM reasoning, alignment, and interaction quality — building models that understand, respond, and collaborate with humans more effectively across modalities.
Industry Track
In industry, I work as a Machine Learning Engineer specializing in Large Language Model systems, building production-grade GenAI applications using vLLM, TensorRT-LLM, RAG pipelines, and agentic workflows. I design scalable LLM inference infrastructure and full-stack solutions across Azure, AWS, and GCP, integrating efficient serving, orchestration, and cloud deployment into real-world AI systems.
vLLM
TensorRT-LLM
LangChain
RAG
Azure
AWS
GCP