Posts

In part one of this blog series , I explored the motivation behind developing a personal recommendation system. The main goals are to learn how recommendation systems work and to build a tool that helps me find interesting blog posts and articles from feeds where only 1 in 20 posts might match my content interests. If you are interested in the technical implementation, the complete codebase is available in this github repository . ...

Every morning, my RSS reader greets me with hundreds of new posts. Tech blogs, indie developers’ journals, photography content - they all compete for attention. While I’ve gotten good at quickly scanning through these feeds , I keep wondering about all the great content I might be missing from sources I’ve had to ignore simply because their signal-to-noise ratio doesn’t justify daily checking. On the other hand, the posts that I shortlist from my RSS feeds and read or listen to, end up on a curated repository of articles that have passed my personal quality threshold, so I have access to a valuable collection of content (on Pocket ) that is relevant to my interests. This made me wonder, can I utilize this data, and create a content recommendation system tailored to my preferences? Can I build a system that would review new posts from feeds where only 1 in 20 posts might match my content priorities, and filter those for me? ...

I purchased a Sony a7ii in 2023, while having zero understanding of the basics of cameras and photography. I recall being particularly confused about the exposure triangle , especially how aperture affects the sharpness and the depth of field of an image. After learning the basics and taking around 3500 photos with that camera, I realized its aging technology was limiting my beginner skills. I had enough experience to be able to research camera models and their features, and I decided to get a Sony a7cii at the beginning of 2024, a decision I’m still happy with today. ...

I’ve been using conversational alpaca or sharegpt formats for fine-tuning LLMs with Axolotl , but it always felt unnecessary to limit the model on a conversational format when the use-case doesn’t require so. I’m currently working on a project to classify pull requests in my company’s code repositories. The model needs to look at the PR title, description, and code changes, then categorize the PR and explain its reasoning. I thought there must be a way to fine-tune these models with any format I see fitting this specific use-case, and sure there is: Template-free Axolotl ...

I was experimenting with some speech-to-text work using OpenAI’s Whisper models today, and transcribing a 15-minute audio file with Whisper tiny model on AWS Lambda (3 vcpu) took 120 seconds. I was curious how faster this could be if I ran the same transcription model on a GPU instance, and with a quick search, modal.com seemed like a nice option to spin up a GPU machine, run the code, and shut down the machine, similar to how AWS Lambda works. ...

TLDR: Here is a simplified diagram of how I keep reading articles, books, and other materials without overwhelming myself or having a never-ending pile of to-read items. In this blog post, I will go into the details of each part of the workflow, the tools I currently use, the alternatives I have tried, why they work for me, and what I can improve in the process. Feel free to use the above table of contents to jump to the section that is more interesting to you. ...

I’ve been using GatsbyJS for publishing my blog posts here, but I wanted to move to another static site generator that is more automation friendly (more on this later). That’s why I decided to migrate this blog to Hugo , which has a very active community and is developed with Go. At first, I was scared of this move, since I don’t know how to code in Go, but to my surprise the whole migration process didn’t require me to write any Go, and everything is handled via yaml, html, and jinja. ...

I’m a big fan of unique charting styles and I avoid using the default matplotlib style whenever possible, as I find it boring and soleless. This preference is not limited to charts, and I also like the hand-drawn styles for fonts and diagrams (excalidraws is a fantastic tool that I use frequently). The hand-drawn style is especially useful when presenting a proof of concept idea. Something very interesting that I’ve recently stumbled upon is an xkcd chart style for matplotlib . This is a fantastic style that I can see myself using frequently going forward. All you need to do is add plt.xkcd() to your existing matplotlib code. ...

Alfred has definitely boosted my productivity since I started using it almost a year ago. Sharing some fragments of my experience using Alfred with a few friends and colleagues made me realize I need to write down a summary of what I find helpful on Alfred, so I can share it with people in future when needed. What’s Alfred Alfred is an app for macOS which boosts your efficiency with hotkeys, keywords, text expansion and more. Search your Mac and the web, and be more productive with custom actions to control your Mac. ...

For a personal project, I was looking for some guidance on how to develop a simple browser extension, but the information on this topic was so fragmanted and difficult to grasp. I then came accross a book named Building Browser Extensions which is focused on the same topic and I found it useful for getting an overall understanding of how to develop a simple extension, but it’s quite long, so I started summarizing the important parts for myself. What follows is just my notes and summaries from this book. ...

I’ve been recently working on survey response data that in addition to aggregatable question types like Likert-scale and multiple-choice questions, includes optional free-text questions. Although we are lucky that thousands of the respondents spend time elaborating on questions and leaving comprehensive free-text responses, getting insights from these text responses is challenging. While investigating how to enrich this text data with proper metadata related to their topics, I came across BERTopic which introduces itself as a topic modeling technique to create clusters allowing for easily interpretable topics. In this post, I’ll explore BERTopic and will go through an example to explain what adjustments worked for me. ...

In my previous blog post I explored using the unstructured python library for loading and parsing documents. As I mentioned in the post, although unstructured seems a very useful library, it has a few issues. Since I’m planning to do a semantic search on the paragraphs and feed the relevant ones to a large language model, the library’s inability to reliably identify headings and paragraphs was a big problem for me. ...

In the world of data, textual data stands out as being particularly complex. It doesn’t fall into neat rows and columns like numerical data does. As a side project, I’m in the process of developing my own personal AI assistant. The objective is to use the data within my notes and documents to answer my questions. The important benefit is all data processing will occure locally on my computer, ensuring that no documents are uploaded to the cloud, and my documents will remain private. ...

Recently, I’ve been working on a side project where I use OpenAI’s text-embedding-ada-002 model to generate vector embeddings for text snippets. While this model is inexpensive, the cost can add up when dealing with thousands or millions of text snippets. Therefore, I decided to explore alternatives, particularly those that would allow me to run similar models locally instead of relying on OpenAI’s API. In this post, I’ll share my experience using the sentence-transformers library for this purpose and discuss the pros and cons. ...

Background For quite a while now, I’ve been using Pocket as my go-to read-it-later app. A few months back, I found myself wanting a solution to listen to my saved articles. This led me to explore the text-to-speech feature in the Reader app. Reader gave me the convenience of linking with my Pocket account, synchronizing its inbox automatically with my saved articles. This system has worked well. I’ve been listening to my saved articles on Reader while continuing to save new articles on Pocket. This lets me easily revisit them later if needed. ...

TL;DR: Whisper.cpp is the fastest when you’re trying to use the large Whisper model on a Mac. For top-quality results with languages other than English, I recommend to ask model to translate into English. About Whisper Whisper is OpenAI’s speech-to-text model and it’s well-known for its impressive results. Although I knew about it for a while, I didn’t get to test its real-world performance until recently. So, I spent a weekend seeing how it could handle converting speeches, in both English and other languages, into text. ...

Recently, I decided to switch my primary password manager to 1Password , using it across my Macbooks and Pixel phones on various browsers, including Firefox and Chrome. One bonus feature that I particularly enjoy is its compatibility with Alfred , an app that helps me be more productive on my Mac. However, the transition was not all smooth sailing. When I transferred my passwords from Firefox and Chrome into 1Password, a couple of things didn’t go quite right. I ended up with: ...

I keep hearing from people that they find it difficult to keep track of what needs to be done. Every one of us has a list of different things to remember, work on, and accomplish. This list includes not only the very well-defined tasks that we call work, but also any Slack messages and emails that we have to reply to, documents that we have to read and review, and vacation time that we need to submit. ...

I use Google Docs every day at work to communicate with various groups of people. It’s a very effortless tool to start writing and share the output with whoever I want once I’m finished. But it definitely lacks a few critical features that hinder my productivity. What I like: Convenience and Collaboration Google Docs makes it easy to write and share documents with others and to invite them to review and collaborate. ...