Writing

100 Days of Gen AI: Day 44

After getting laid out with strep throat, I’m back. Today I want to explore what I can do with Replit, a code generation and execution platform. Code execution has been...

100 Days of Gen AI: Day 43

An idea I’ve wanted to explore is how to turn a paper form into an API. The basics would be to take a photo of the form, use OCR to...

100 Days of Gen AI: Day 42

I want to try fine-tuning locally on my macbook M4. Unfortunately I can’t use Unsloth since a mac doesn’t meet the Unsloth system requirements. I’m sure I can still...

100 Days of Gen AI: Day 41

Today I fine-tuned an LLM model. I started out with this unsloth tutorial, where we fine-tune atop llama 3. The tutorial lays everything out in a relatively straightforward way....

100 Days of Gen AI: Day 40

Let’s try running the new DeepSeek-R1 model locally! The model is too large for my M4 macbook, but I will try to run a much smaller distilled model. I’m using...

100 Days of Gen AI: Day 39

Today I’m calling the Siri Playlist Actions project done. It was a fun week-long project. You can see it running here: https://spotify.woolgathering.io/ Find the code on Github here:...

100 Days of Gen AI: Day 38

I’ve fully refactored the service into serverless functions, deployed to Vercel, and now have the site hosted at https://spotify.woolgathering.io/! This allows any user to use the service, without running...

100 Days of Gen AI: Day 37

Today I got the remove song from current playlist endpoint working. I also found and fixed several bugs throughout the other endpoints. Next I need to refactor how the code...

100 Days of Gen AI: Day 36

Today I kept working on the Apple Siri shortcut. I updated how tokens are managed in Redis such that a subsequent login will reuse the existing API token rather than...

100 Days of Gen AI: Day 35

Today I kept working on the Apple Siri shortcut. I used ChatGPT to help me setup a database to persist the spotify session tokens. At first I used an in-memory...