Open weight and closed models differ in one powerful idea. Who holds the keys to the engine? Open weight gives you full access to the intern…
When you think about evaluating an LLM, the helpful way to see it is this. Automated metrics give you quick signals while humans bring the d…
What is an agentic AI system? An agentic AI system is best understood as an AI that doesn't wait for every instruction. You give it a goal, …
Think of a context window as the space in your LLM's short-term memory. It's the amount of text it can actively pay attention to at one time…
What are prompt engineering best practices? Prompt engineering works best when you guide the model the same way you guide a teammate who wan…
When you think about evaluating an LLM, the helpful way to see it is this. Automated metrics give you quick signals while humans bring the d…
Open weight and closed models differ in one powerful idea. Who holds the keys to the engine? Open weight gives you full access to the intern…
How to optimize inference latency for large language models. Optimizing inference latency is a bit like tuning a long road trip. You start b…
Think of a context window as the space in your LLM's short-term memory. It's the amount of text it can actively pay attention to at one time…
What's the role of temperature in top P? Before we get into the details, here's the simple idea behind this topic. Temperature controls how …
What are embeddings and how are they used in rag? Embeddings are simply numbers that capture the meaning of text. Think of them as a map whe…
Positional encoding simply gives transformers a sense of order. Without it, a transformer sees all words at once with no clue who came first…
What are hallucinations in LLMs? Hallucinations happen when an AI gives an answer with full confidence, even though the information is false…
How to build a rag system end to end? Rag stands for retrieval augmented generation and the simplest way to think about it is the model chec…
Lurai is a technique called low rank adaptation and the key idea is surprisingly simple. Instead of updating all the weights of a massive mo…
How does tokenization work in LLMs and why is it important? Tokenization is simply how LLMs break text into smaller pieces they can understa…
Fine-tuning, instruction tuning, and RLHF build on each other. So, let's walk through them slowly and clearly. Fine-tuning is the first step…
Let's start with the big picture. The transformer is an architecture that processes all tokens in a sequence at once instead of step by step…
Let's start with the big picture because these three models make sense once you see how each one reads text. GPT moves left to right and pre…
Want transcripts for the whole Peetha Academy channel?
Get All Peetha Academy Transcripts →If you found this helpful, consider leaving a tip!