About
Activity
-
Physical AI Days return to #NVIDIAGTC —now in Taipei. Join curated sessions covering robotics, industrial AI, digital twins, and simulation—all…
Physical AI Days return to #NVIDIAGTC —now in Taipei. Join curated sessions covering robotics, industrial AI, digital twins, and simulation—all…
Liked by Ahmad Qardahji
Experience & Education
-
Intuit
View Ahmad’s full experience
See their title, tenure and more.
or
Licenses & Certifications
-
-
Soft Skills (Communication and Leadership)
German - Jordanian University
Issued -
-
Volunteer Experience
-
founder of C Social Enterprise (German - Jordanian University)
(German - Jordanian University)
Courses
-
Algorithms and Data Structures
-
-
Artificial Intelligence
-
-
Component-Based Computing
-
-
Computer Aided Engineering Drawing
-
-
Computer Architecture and Organization
-
-
Computer Networks(MPLS)
-
-
Data Communication
-
-
Database Design
-
-
Database Management Systems
-
-
Digital Systems
-
-
Electrical Circuits, Analog Electronics, Digital Electronics
-
-
Electromagnetic Fields and Waves
-
-
Emotional intelligence and leadership
-
-
Fundamentals of Analog Electronics
-
-
Fundamentals of Digital Electronics
-
-
Graphical User Interfaces
-
-
Information Security
-
-
Intercultural Communication
-
-
Microprocessor and Microcomputer Systems
-
-
Network Management Systems
-
-
Networks
-
-
OOP (Java)
-
-
Operating Systems(Linux, Unix, Win, Solaris)
-
-
Project management
-
-
Real Time Systems
-
-
Signals and Systems
-
-
Social Entrepreneurship and Enterprises
-
-
Soft Skills (Communication and Leadership)
-
-
Software Engineering
-
-
Systems Programming
-
Honors & Awards
-
One year full scholarship offered from the DAAD for an exchange year in Germany
German - Jordanian University
Scholarship that covers traveling and living expenses in Germany offered for the first on the school of Computer Engineering and Information Technology
-
Scholarship from the DAAD for two semesters for excellence in German language
German - Jordanian University
-
Half scholarship offered from Drexel University, Philadelphia, Pennsylvania
Drexel University
Scholarship for top notch students selected from international universities
-
DAAD scholarship for German language course in Munich
German - Jordanian University
Scholarship for one month offered for students with outstanding German language abilities.
-
3rd place on Jordan in Scientific research
Ministry of Education, Jordan
-
Award for distinguished academic performace
German - Jordanian University
This award was given to distinguished students based on their excellent&outstanding records at the German - Jordanian University. First of School of Computer Engineering and Information Technology
Languages
-
Arabic
Native or bilingual proficiency
-
English
Full professional proficiency
-
German
Professional working proficiency
Other similar profiles
Explore more posts
-
Apala Guha
Ruokai Yin, our summer intern from Yale to Microsoft Azure, worked with Xuan Zuo, @Sattwik Deb Mishra, Hokchhay T. and Preyas Shah and me. We published a paper based on his project: https://lnkd.in/eHuG3BSt Summary of paper: Finding optimal sharding is hard especially for large LLMs. A large number of devices have to be divided into TP, PP, EP dimensions. For TP, we have to decide the tensor dimension e.g. head, hidden size to shard for every op. Additionally, TP can be one-dimensional/two-dimensional etc. This makes for a large sharding space that is hard to cover exhaustively. SoTA methods use heuristics to select sharding e.g. GSPMD, Sequence Parallelism, Helix etc. There are significant time gaps, months or years, between the discovery of each of these methods. We want to fast-forward the process of discovery by 1) using a formal representation of the sharding space, and, 2) using reinforcement learning to search the sharding space. This paper demonstrates that our method is 3.5x better than random search, and on par with hand-tuned heuristics when looking only at 1D sharding and without allowing sequence parallelism. We aim to extend the work to multi-dimensional sharding as well as sequence parallelism. Beyond that, we wish to combine the problem of sharding with fusion, tiling and collectives as well.
-
Douglas Pan
One of the biggest barriers to trusting AI results, beyond hallucination, is nondeterminism: the inability of Large Language Models (LLMs) to consistently produce identical outputs for identical inputs or prompts. This raises concerns in the tech community as reproducibility is a cornerstone of scientific progress. In the paper "Defeating Nondeterminism in LLM Inference” (Thinking Machines Lab: Connectionism, Sep 2025), Horace He and Thinking Machines Lab identify this nondeterminism to be not just due to randomness in GPU math, but to variation in batch size under load. This load arises from parallel user queries that may lead to different reduction orders in operations like matrix multiplication, normalization, and attention. By implementing batch-invariant kernels and using consistent reduction strategies, such as fixed tile sizes for matrix multiplication and fixed split sizes for attention, they demonstrate that nondeterminism can be eliminated with some performance tradeoffs. Therefore, LLM inference nondeterminism, previously viewed as an inherent problem, does have an engineering solution. https://lnkd.in/gvmWVas8
2 Comments
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content