Different Forms of Paper Novelty Receive Uneven Recognition
An NLP/ML study using language-derived novelty measures to analyze scholarly recognition and public attention.
沈悠然
ML SystemsLLM InfrastructureHPC
Computer Science student at Texas A&M University building efficient training, inference, and agentic systems. My work spans distributed LLM serving, graph-based retrieval and reasoning, and real-time autonomous systems.
Texas A&M Department of Computer Science & Engineering
Continuing as a teaching assistant in Fall 2026 for the CSCE 120 Go section with Dr. Philip Ritchey.
SKY Lab, Texas A&M University
Working with Dr. Yu Zhang on Graph RAG for graph-structured retrieval and reasoning. Exploring agent planning, tool-calling validation, controllable reasoning, reinforcement-learning failure recovery, and embedding-based semantic analysis.
TAMU Supercomputing Team (Gig ’Em Bytes)
Optimized MLPerf LLM training and inference on Grace and ACES clusters using TensorRT, Triton, vLLM, Ray, and MaxText. Worked on multi-node latency and throughput, telemetry pipelines, concurrent post-training, and ACTS tracking-software performance. Our team placed 5th globally at SC25 SCC.
Texas A&M Department of Computer Science & Engineering
Mentored 60+ students weekly in data structures, algorithms, operating systems, and C++. Created review videos and graded problem sets for discrete mathematics with Dr. Sandeep Kumar, alongside competitive-programming coaching.
Built an on-premise RAG system with Qwen-7B and QAnything for document question answering and permissioned retrieval, reducing internal document retrieval time by 90% for more than 100 employees. Also developed FastAPI safety-monitoring services and real-time data pipelines across corporate facilities.
Developed real-time machine-learning quality control for calibration and anomaly detection. Co-developed a TensorFlow autoencoder for medical-image denoising, with a focus on reducing server CPU utilization.
An NLP/ML study using language-derived novelty measures to analyze scholarly recognition and public attention.
Bachelor of Science in Computer Science
Embedded perception with Texas Aimbots, Mars rover autonomy with KARURA, agricultural robotics, and a lunar prospecting design. Explore the work alongside team videos, television coverage, and official competition records.
A real-time embedded perception stack for competition robots.
Risk-aware autonomous navigation for a Mars rover platform.
Computer vision and autonomous navigation for agricultural pest control.
Large-scale lunar crater prospecting for the 2024 RASC-AL design challenge.
Fun fact: I also design websites for fun.
"君子矜而不争,群而不党"与求同存异之殊异
这是本博客测试与发布的第一篇文章。使用Hugo和PaperMod主题创建。
My first post in English — welcome to the English version of my blog.