Ph.D. candidate in Computer Science · Iowa State University
I build tools that make machine learning systems easier to debug, test, and trust.
I am Anwar Hossain Zahid, advised by Prof. Wei Le. My work sits at the intersection of software engineering, machine learning reliability, and numerical analysis across model versions, compilers, and GPU platforms.
Current Focus
Reliability for ML systems that change over time
Differential ML Debugging
Finding behavioral regressions across model versions using differential testing, invariant learning, and targeted analysis.
Numerical Stability
Detecting silent failures caused by unstable floating-point behavior in ML applications and scientific workloads.
Cross-Platform GPU Testing
Comparing NVIDIA and AMD GPU computations to expose portability and reproducibility issues in high-performance code.
Selected Work
Recent research and engineering
I have worked on Soft Assertions for ML numerical instability, GPU numerical testing at Lawrence Livermore National Laboratory, and LLM evaluation for social-good applications.
Explore publications and projectsWriting
Notes from research and practice
What I Want This Blog to Become
I want this blog to be a working notebook for the problems I keep returning to: machine learning reliability, numerical instabi...
Jul 8, 2025Automatically Detecting Numerical Instability in ML via Soft Assertions
Machine learning (ML) models run on massive datasets and often perform billions of floating-point calculations. But here’s the ...
Jul 7, 2025Testing GPU Numerics: Finding Numerical Differences Between NVIDIA and AMD GPUs
When you run the same GPU program on an NVIDIA GPU and an AMD GPU, you might expect identical results. Surprisingly, that’s not...
Open To
Research collaboration, internships, and engineering conversations
I am especially interested in ML reliability, debugging infrastructure, numerical correctness, AI safety, and systems that connect research ideas with working software.