These videos cover configuration, use and thoughts around the NVIDIA DGX Spark platform
5 videos
18:10
This video walks through the concepts and implementation of GraphRAG, a method that enhances language model retrieval by incorporating graph databases. Viewers will learn how to build a full-stack GraphRAG application, including setting up the backend, integrating a graph database, and connecting it with an LLM for a more accurate and explainable AI experience.
![Vectorizing Images with LLMs [Full Stack Deep Dive]](https://i.ytimg.com/vi/BQqmkyBAU_I/maxresdefault.jpg)
13:04
This video walks through building a complete end-to-end image vectorization and similarity search system using modern AI tools, including multimodal language models and GPU-powered infrastructure. You'll learn how to generate image embeddings with an OpenAI model, store vectors efficiently with ChromaDB, and set up LLM microservices for semantic search, culminating in a live demo of the final solution.
37:19
This video walks through the full workflow of fine-tuning the Phi-4-Mini-Instruct model using LoRA and PyTorch, covering how to code a fine-tuning script, format custom training data, run fine-tuning jobs in a Docker container, merge fine-tuned adapters, quantize the model for efficient local device operation, and test the fine-tuned model. Viewers will gain hands-on experience with these techniques applicable to any CUDA-based environment.
5:10
This video walks through the process of setting up OpenAI’s GPT-OSS-20B locally using Docker and Open WebUI, allowing viewers to experiment with a large language model on their own hardware. You'll learn how to create a Docker container, set up NVIDIA Sync for remote access, and run the model on various types of machines, including desktop-class systems like the NVIDIA DGX Spark.
5:47
This video walks through the NVIDIA DGX Spark, detailing its hardware specifications and value proposition, while also exploring who this data-center-grade AI lab is suitable for and who might find it less appropriate.