Recent content in Home on syaffers.xyz
This is the C.H.I.P. I took this photo on July 25, 2016, when it first arrived at my college dorm room. It will turn 10 years old next month under my possession. Trip down memory lane The C.H.I.P. was (is?) a $9 computer built by Next Thing Co. (NTC from now on), a company that is now defunct. I recall the hype for the C.H.I.P. around the Twitter-verse (being on the radar for hobbyist hardware) at…
What would run Qwen3-0.6B faster on CPU: a Surface Go from 2018 or a Raspberry Pi 5 from 2023 ? What if you knew the respective CPUs: an Intel Pentium 4415Y versus an ARM Cortex A76 ; did your guess change? The 4415Y processor is clocked at 1.6 GHz and the A76 is around 3.0 GHz . How much faster would it run? 2x? 5x? Answer: it’s the Raspberry Pi 5 and by a difference of at least 10x.
New tricks for an old dog I recently revived my Raspberry Pi 2 Model B (rev 1.1) . It’s a decade-old piece of hardware to be sure, but I have a breakout garden HAT (not sponsored, nor affiliated) with ambient sensors that I hadn’t use in a few years. So I thought: let’s get Claude’s juices flowing. Let it draw inspiration from the Pimoroni Python libraries and build a fast…
For the answer to “How do I serve Qwen3-VL-Embedding-2B (or 8B)?” see the last section . The rest of this post is how I got there. I needed to deploy Qwen3-VL-Embedding-8B for a customer recently. I knew of Qwen’s models: very popular, open-weights, and competitively performant. Qwen’s models play nicely with model serving frameworks like vLLM and SGLang due to their…
My Hugo blog is coming together swimmingly. I’m comfortable in Markdown and Hugo makes it quick to build; big fan. What I’m sometimes not a big fan of: is my idiosyncratic desire to be technical when I don’t need to be. Let me explain: I like VPSes It’s your little computer-away-from-computer. It’s accessible, cheap (if you know what you’re getting), and you…
I left Medium for my own blog, bringing the three chunky posts I have on there with me. It has been a while since I’ve posted there, partly due to my busy schedules. What’s more pertinent are my sentiments towards Medium. Content quality has nosedived; thanks algorithms AI slop and clickbait: the two most prominent appearances on my feed. These contain the most uninspired, sophomoric…
A few weekends ago, I wanted to brush up on my clustering techniques. I recall — from an image processing class back in college — that you can reduce the number of unique colors in an image down to a few colors while keeping the important ones. This is achieved by doing k-means clustering on the RGB values of the pixels in an image. The number of colors that you want in your reduced palette is the…
Once in a while, you get to work with the underdog of the data world: time series (images and natural language have been in the limelight a lot, recently!). I’ve been lucky (or unlucky) enough to have had to work on time series data for the most part of last year. Among my adventures in the wacky world of 1-dimensional machine learning, I found some very useful vectorization that allowed me…
PyTorch has been around my circles as of late and I had to try it out despite being comfortable with Keras and TensorFlow for a while. Surprisingly, I found it quite refreshing and likable, especially as PyTorch features a Pythonic API, a more opinionated programming pattern and a good set of built-in utility functions. One that I enjoy particularly well is the ability to easily craft a custom…
You have just built an awesome convolutional neural network which takes images and spits out what the object in the image is (to some degree of certainty). You want to share it with your peers around the world to get some feedback on how well it is performing while, at the same time, collect training data to train your model more. While it can be tempting to create a desktop application, this does…
I’m Syafiq Kamarul Azman, a Software Engineer (Python) from KL, Malaysia 🇲🇾. I’m currently living and working in AD, UAE 🇦🇪. I’m passionate about: Machine Learning & AI - from neural networks, game-playing AI, MLOps, even LLMs, Visualization - seeing complexity and explaining it, Open Source - learning new tools and contributing back, Raspberry Pi - and other SBCs and…
A collection of past projects, publications, and talks. Talks The Assembly Palette Transfer Talk 🎬 Presentation slides. PPTX file 📹 Livestream video. YouTube recorded livestream 📔 Palette transfer notebook. Google Colab The Assembly Transfer Learning Talk 🎬 Presentation slides. Google slides 📹 Livestream video. YouTube recorded livestream 📔 Transfer learning: machine learning notebook.…