12 LLM Projects For All Levels

LLMs Projects for Beginners 

In the beginner projects, you will learn how to use the OpenAI API to fine-tune models, generate responses, build AI assistants with multimodal capabilities, and serve API endpoints of your application using FastAPI.

Prerequisites: An understanding of Python is necessary. 

1. Fine-Tuning OpenAI's Models

Project link: Fine-Tuning OpenAI's GPT-4

Fine-tuning GPT-4 on a custom dataset has become quite easy. You don't need expensive GPUs, to install software, or to write Python code. The only thing you need to do is upload the dataset and run the fine-tuning OpenAI API. Simple. 

After fine-tuning is completed, you can use this model in your AI application, similar to any other model using the OpenAI API. 

In this project, you will learn about the OpenAI API and how to modify and upload the dataset. In the end, you can access the fine-tuned model and generate the customized response.

Source: Fine-Tuning OpenAI's GPT-4

2. Building an AI Assistant for Data Science with Multimodal Capabilities

Project link: Creating AI Assistants with GPT-4o

In this project, you will be creating your own AI assistant similar to ChatGPT. You will learn to use the OpenAI API and GPT-4o to create a specialized AI assistant tailored for data science tasks. This includes providing custom instructions, handling files and data, and using multimodal capabilities. 

The fun part is that the whole project only requires you to understand various functions of the OpenAI Python API. With a few lines of code, you can build your own automated AI that will process the data for you and present you with the analysis. 

Source: Creating AI Assistants with GPT-4o

3. Serving an LLM Application as an API endpoint using FastAPI

Project link: Serving an LLM application as an API endpoint using FastAPI in Python

What is the next step after building your application using the OpenAI API? You will serve and deploy it to a server. This means creating a REST API application that anyone can use to integrate your AI application into their systems. 

In this project, you will learn about FastAPI and how to build and serve AI applications. You will also learn how API works and how to test it locally. 

Source: Serving an LLM application as an API endpoint using FastAPI in Python

Intermediate LLMs Projects

In the intermediate LLM projects, you will learn to integrate various LLM APIs like Groq, OpenAI, and Cohere with frameworks such as LlamaIndex and LangChain. You will learn to build context-aware applications and connect multiple external sources to generate highly relevant and accurate responses.

Prerequisites: Experience with Python and an understanding of how AI agents work. 

4. Building a Context-aware ChatPDF Application

Project link: Groq LPU Inference Engine Tutorial

The Groq API is similar to the OpenAI API but offers a speed of almost 500 tokens per second compared to OpenAI's 45 tokens per second, providing real-time responses. 

In this project, you will learn about the Groq Python API and how to use it to build AI applications. You will explore concepts such as vector stores, indexing, embeddings, LLMs, chat engines, and memory buffers. 

By using these components, you will build a context-aware RAG (Retrieval-Augmented Generation) application that can generate relevant responses using PDF files.

Source: Groq LPU Inference Engine Tutorial

5. Building a DuckDB SQL Query Engine Using an LLM

Project link: DuckDB Tutorial: Building AI Projects

DuckDB is a modern, high-performance, in-memory analytical database management system (DBMS) that can be used as a vector store and a SQL query engine. 

In this project, you will learn about the LlamaIndex framework to build an RAG application using DuckDB as a vector store. Additionally, you will build a DuckDB SQL query engine that converts natural language into SQL queries and generates responses that can be used as context for the prompt. It is a fun project, and you will learn about a lot of LlamaIndex functions and agents. 

Source: DuckDB Tutorial: Building AI Projects

LLM Projects for Final-Year Students

Final-year students have the opportunity to build proper LLM applications. Therefore, most projects mentioned in this section require knowledge about fine-tuning, LLM architectures, multimodal applications, AI agents, and, eventually, deploying the application on a server.

Prerequisites: Good understanding of Python language, LLMs, AI agents, and model fine-tuning.

6. Building Multimodal AI Applications with LangChain & the OpenAI API

Project link: Building Multimodal AI Applications with LangChain & the OpenAI API

Generative AI makes it easy to move beyond simply working with text. We can now combine the power of text and audio AI models to build a bot that answers questions about YouTube videos. 

In this project, you will learn how to transcribe YouTube video content with the Whisper speech-to-text AI and then use GPT to ask questions about the content. You will also learn about LangChain and various OpenAI functions. 

The project includes a video guide and code-along sessions for you to follow and build the application with the instructor.

Take the Developing LLM Applications with LangChain course to learn more about its features through interactive exercises, slides, quizzes, video tutorials,  and assessment questions.

Source: Building Multimodal AI Applications with LangChain & the OpenAI API

7. Fine-Tuning Llama 3 and Using It Locally

Project link: Fine-Tuning Llama 3 and Using It Locally: A Step-by-Step Guide

Fine-tuning using the OpenAI API is easy, but it has limitations. It can only fine-tune the OpenAI model, and you are sacrificing privacy while also paying more compared to fine-tuning on your own machine. Ultimately, it all comes down to cost and security. 

In this project, you will learn to use the Transformers framework to fine-tune the Llama 3 model on patient and doctor conversations. Additionally, you will learn to merge the base model with the fine-tuned adapter, convert it into GGUF format, quantize it, and use it locally with a Jan application. 

By completing this project, you will gain knowledge on LLMs, fine-tuning, quantization, and various methods to optimize memory and compute usage.

Source: Fine-Tuning Llama 3 and Using It Locally

8. Build a Multiple-step AI Agent Using the LangChain and Cohere API

Project link: Cohere Command R+: A Complete Step-by-Step Tutorial

Cohere API is similar to OpenAI API and offers many features compared to other LLM frameworks. In this project, you will learn the core features of Cohere API and use them to build a multi-step AI agent. 

You will also master the LangChain framework and chaining together various AI agents. 

In the end, you will build an AI application that will take the user's query to search the web and generate Python code. Then, it will use Python REPL to execute the code and return the user's requested visualization. 

Source: Cohere Command R+: A Complete Step-by-Step Tutorial

9. Deploying LLM Applications with LangServe

Project link: Deploying LLM Applications with LangServe

In this project, you will learn how to deploy LLM applications using LangServe. You will start by setting up LangServe, creating a server using FastAPI, testing the server locally, and then deploying the application. You will also learn to monitor LLM's performance using the Prometheus framework and then use Grafana to visualize the model performance in the dashboard. 

There is a lot to learn, and it will help you get a job in the LLMOps field, which is growing in demand. 

Source: Deploying LLM Applications with LangServe

Advanced LLMs Projects

The advanced LLM projects are designed for experts who want to build production-ready applications and improve already deployed LLM applications. By working on these projects, you will learn to infer and fine-tune on TPUs instead of GPUs for faster performance, automate taxonomy generation and text classification, and build AI applications in the Microsoft Azure Cloud.

Prerequisites: Good understanding of Python, LLMs, model fine-tuning, TPUs, and cloud computing.

10. Fine-Tune and Run Inference on Google's Gemma Model Using TPUs

Project link: Fine-Tune and Run Inference on Google's Gemma Model with TPU

Fine-tuning and running model inference on TPUs come with their own set of issues and limitations, but in return, you get 4x the speed compared to a multi-GPU setup. 

In this project, you will learn about CPUs, GPUs, and TPUs, and how you can use TPUs to generate faster responses and fine-tune models in a quarter of the time. You will also learn about distributed fine-tuning and inference of models with TPUs, achieved through model parallelism, which distributes a single model's weights across multiple devices, enabling horizontal scaling and accelerating training.

Source: Fine-Tune and Run Inference on Google's Gemma Model with TPU

11. Build a TNT-LLM Application

Project link: GPT-4o and LangGraph Tutorial: Build a TNT-LLM Application

Microsoft's TNT-LLM is a cutting-edge system designed to automate taxonomy generation and text classification by leveraging the power of large language models (LLMs), such as GPT-4, to enhance efficiency and accuracy with minimal human intervention. In this project, you will implement TNT-LLM using LangGraph, LangChain, and OpenAI API. 

You will define graph state class, load datasets, summarize documents, create minibatches, generate and update taxonomy, review it, orchestrate the TNT-LLM pipeline with StateGraph, and finally cluster and display the taxonomy.

Source: GPT-4o and LangGraph Tutorial

12. Building an AI Application Using Azure Cloud

Project link: Using GPT on Azure

Gaining cloud computing experience is a crucial skill that will land you a position in any LLM-related field. Mastering the ability to build, automate, deploy, and maintain an LLM pipeline on Azure cloud will earn you the title of an expert in LLMs. 

In this project, you will learn how to set up a pay-as-you-go instance of OpenAI and build simple applications that demonstrate the power of LLMs and Generative AI with your data. You will also use a vector store to build an RAG application and reduce hallucination in the AI application. 

Source: Using GPT on Azure

Final Thoughts 

Working through the projects mentioned above will provide you with enough experience to build your own product or even secure a job in the LLM field. The key to success is not just coding or fine-tuning but documentation and sharing your work on social media to attract attention. Ensure your project is well-explained and user-friendly, including a demo or usability guide within the documentation.

In this blog, we reviewed 12 LLM projects that can help you land your dream job as an LLM Scientist. If you are interested in learning more about LLMs, we highly recommend exploring the Developing Large Language Models skill track. It will teach the latest techniques for developing cutting-edge language models, mastering deep learning with PyTorch, building your own transformer model from scratch, and fine-tuning pre-trained LLMs from Hugging Face.