LLM

Clearing out a year-old Obsidian folder on model fine-tuning made the shift obvious: from chasing model capability to chasing business results through workflow and RAG.

Deploying the open-source RAGFlow project, testing it against real documents, and wiring it up to local Ollama models to keep costs at zero.

Running ten locally-hosted models on consumer GPU hardware to see which ones best summarize customer-service calls after speech-to-text.

What an ‘agent’ actually means in the context of large language models, and a worked example in a customer-service setting.

A one-line fix for a CUBLAS_STATUS_NOT_INITIALIZED crash: just update Ollama.

After AMD’s ROCm ecosystem let me down for local fine-tuning, free Colab GPUs turned out to be the pragmatic way to fine-tune Llama 3 for free.

Pairing a consumer GPU running local Llama 3 with Open WebUI and One API to combine local and remote models behind one interface.

Wrapping Cloudflare’s free Workers AI models (including Qwen) behind an OpenAI-compatible endpoint, so existing front-ends don’t need to change a line of code.

A quick note on a comparison table of large-model offerings across major cloud providers.

Some hands-on experiments with un-tuned open models, and a few thoughts on where the real value in LLM products actually sits.