Recent content on AmineDiro
⏲️ Estimated reading time ~45min. Flash Attention: From Theory to Implementation Flash Attention has become one of the most impactful optimizations in modern deep learning. Since the original paper was published in 2022, we’ve seen four major versions—each squeezing more performance out of increasingly powerful hardware. But here’s the thing: reading papers is one thing, understanding…
I recently listened to the rust in production podcast with the founder of tembo.io. The guest and host talked about their PGMQ rust project, a job queue implemented in postgres. They went into some details of how they build this using the amazing pgrx rust lib for writing Rust postgres extension. I wouldn’t really recommend using Postgres as a message broker nor a task queue for obvious…
A few years ago, I was really into Raspberry Pis. I think they are amazing pieces of hardware to own if you love tinkering and learning new stuff. By stuff, I mean: software, embedded systems, kernel development, self-hosting, networks, clusters, robotics, and more. I still own quite a few. Raspberry Pis are a great educational tool, they only require swapping out a single SD card to boot up into…
A few years ago, I was really into Raspberry Pis. I think they are amazing pieces of hardware to own if you love tinkering and learning new stuff. By stuff, I mean: software, embedded systems, kernel development, self-hosting, networks, clusters, robotics, and more. I still own quite a few. Raspberry Pis are a great educational tool, they only require swapping out a single SD card to boot up into…
A few weeks ago, I watched my daily primetime video and it featured the billion-row challenge. I first chugged it off because I immediately saw the J-word ( I’ll say it this time … Java 🤮 ) sprinkled around. Then I nerd-sniped myself into practicing some Rust and trying to dunk on those GC 1GB VM-needing languages. In this blog post, I’ll walk through how I optimized my…
A few months ago, I attended a Rust meet-up presentation focused on webassembly presented by the French startup Rayon. I read a little about the technology and watched two or three talk about it. Coincidentally, I also took an interest in the new wgpu graphic API presented by Google a little bit over 7 months ago, the API was particularly interesting to me because they introduced compute shaders.…
Sorry in advance for the catchy title, serving embedding is usually bottlenecked by computing it first ie running the model on the input. That is not what I’m going to dive into below. This is more of a hand’s on experimentations article where I try different methods on client and server to send and receive numpy arrays. In this article, I’ll try to compare different methods of serving numpy…
In his infamous video “Clean” Code, Horrible Performance, the legendary Casey Muratori showed how trying to be cute with your code and introducing unnecessary indirection can hurt performance. He compared the “clean” code way of structuring your classes in an “OOP” style, using class hierarchy, virtual functions, and all the hoopla. He then showed how writing a…
As a data scientist, I see a lot of ML/AI projects boasting about what amazing things their library can do using ‘pip install’ and 3 lines of Python. In parallel, there is this insatiable need for data scientists to have libraries that can do every single little computer science thing to “offload” that mental burden of actually learning how the actual system works and quickly do the…
I started working on a distributed task queue library a few months back. Now, what is a distributed task queue library? Let’s say that you theoretically want to process a huge number of documents. Each document is composed of a variable number of pages. Processing each task takes a variable amount of time. This is a classic embarrassingly parallel workload. Each page can be processed in parallel…
This article is the third and last article from the should I get a data science degree series. In the first two articles (check out part I and part II), I gave my perspective on self-taught data science vs classically trained. I tried to answer this question from my perspective i.e through the lens that my experience of first self-learning through MOOCs and various online resources, then actually…
This article is the second part of a three-part series of articles on self-taught vs “classically” taught data scientists from my perspective. In the first part, I wrote about my journey from hating my job to starting a self-taught data science journey. I gave some brief insight on the online resources that helped me along the way, I also took some time to explain my background and why I’m writing…
Bro: Are you stupid? definitely go to school to study data science. Other Bro: Yeaah…. maybe not, you should probably avoid wasting moneys’ and learn everything online. The internet ever heard of it, bro ?. Bro: Weeeeelll, you should learn some math before calling yourself a data Scientist though, no? And there they went for ages debating the fall of the educational system and the internet MOOC…
“Nous sommes en guerre” le verbatim du discours d’Emmanuel Macron, répété plus de 6 fois ce lundi 16 Mars 2020. Dans un monde où la panique et la propagande ont remplacées le discernement, la raison et l’expertise, je me tourne vers la Data comme lumière qui pourrait éclaircir la situation et nous donner les moyens d’évaluer la gravité du confinement total en France. Sommes nous vraiment en guerre…
As a longtime martial artist and an MMA fan, I saw the progression of the sport known as mixed martial arts from a competition between two old fighting styles to a real sport, where competitors took what works from each martial art, tailored it to their fighting abilities to serve them best when facing other trained killers. From the first UFCs where fighters where strict boxers, wrestlers, BJJ…
🇲🇦 I’m a Moroccan Machine Learning engineer and data scientist. 💪 I’m the co-founder of Fitroulette App,a webRTC based application for enabling remote fitness sessions 🤖 I have experience training and deploying machine learning models in both Cloud and HPC environments. I have built and optimized pipelines to process terabytes of data. I like to make things go ‘brrr!’ 💬 You can…