About Me
Welcome!
I’m Ahmad Saeed Zaidi, an Artificial Intelligence undergraduate (6th semester) at the Ghulam Ishaq Khan Institute of Engineering and Technology (GIKI).
I specialize in data pipelines, low-level systems, and machine learning. I am currently seeking internships or full-time roles where I can build high-performance systems or robust data infrastructure.
Data Hoarding
My absolute favorite thing to do is scrape data. If it exists, I want it. I enjoy building automated MLOps pipelines to extract, transform, and archive massive datasets. My recent targets include:
- Deeply annotated CUDA kernels from top GitHub repositories.
- Multimodal video datasets and analytics/statistics from YouTube.
- Granular, historical weather data.
Systems, CUDA, & Open Source
I spend a significant amount of my time reading AI research papers and studying CUDA memory hierarchies and kernel optimizations.
- Currently Reading: The source code of vLLM, llm.c, flashinfer and TensorRT-LLM, as well as the DeepSeek-v4 research paper.
- Open PR: PR 3460 on flashinfer, improving mhc_post kernel performance for sm90 architecture, profiled using cuobjdump file on modal h100.
- Open Source Goals: I aspire to make meaningful contributions to the open-source community, specifically targeting VLC and various high-performance CUDA libraries.
- Core Stack: C++, Python, CUDA, MLOps.
Beyond the Screen
When I’m not working, I am usually:
- Speed cubing
- Solving math problems
- Folding origami
- Playing Tabletop RPGs (and running Spright/Kewl-Tune in Yu-Gi-Oh!)
- Reading manga
- Cooking or working out (mental and physical well-being is important!)
Contact
The fastest way to reach me is on Discord: diablo.sama