Teknosys learning library

1. Define the task and choose a development route

Decide whether you need a classifier, a summarizer, a conversational assistant, or a model for another task. Define the languages, latency, hardware budget, acceptable failure cases, and evaluation dataset first.

• Prompting: guide an existing model without changing weights.
• RAG: retrieve relevant documents into the prompt for changing knowledge.
• Fine-tuning: update weights or train adapters to learn task-specific behavior.
• Pretraining: learn from a large corpus, usually requiring substantially more data and compute.

A good first project is adapting a small pretrained model and comparing it against its unchanged baseline.

Educational material. Verify examples against current documentation and applicable licenses. Report a correction.