Coding

All of these projects are accessible on my GitHub. Additionally, for the latest ones, I've provided comprehensive video walkthroughs on my YouTube channel, covering all the details.
Farzad Roozitalab

Large Language Models (LLMs) and generative AI:

[Github Link] HUMAIN (Advanced Multimodal Chatbot): A multimodal chatbot that is designed using OpenAI, HuggingFace, Langchain, and Gradio, offering these capabilities:

  • Can act as an AI assistant (similar to ChatGPT)
  • Has 3 RAG capabilities: RAG with processed docs, upload docs, and websites
  • Can summarize documents and websites
  • Can automatically search the web using the DuckDuckGo search engine based on the user's query, and provide a proper response.
  • Can understand text, voice, and image.
  • Has built-in memory for the GPT models.
Models:
    GPT 3.5: OpenAI || text-embedding-ada-002: OpenAI || llava-hf/llava-v1.6-mistral-7b-hf: HuggingFace || stabilityai/stable-diffusion-xl-base-1.0 : HuggingFace || openai/whisper-base.en: HuggingFace
**Youtube Link**

[Github Link] LLM Fine-tuning: Fine-tuning 3 LLMs on dataset of a fictional company and designing a chatbot using the fine-tuned model. The project covers the whole pipeline from the first steps of data processing upto desinging the chatbot webapp. **Youtube Link**

[Github Link] RAG-GPT: Fostering Robust AI Conversations with retrieval Augmented generation By Integrating OpenAI GPT Model, Langchain, ChromaDB, and Gradio. RAG-GPT is your interactive document companion, facilitating seamless conversations with your documents, including both PDFs and Docs.
The chatbot offers versatile usage through three distinct methods:

  • Offline Documents: Engage with documents that you've pre-processed and vectorized. These documents can be seamlessly integrated into your chat sessions.
  • Real-time Uploads: Easily upload documents during your chat sessions, allowing the chatbot to process and respond to the content on-the-fly.
  • Summarization Requests: Request the chatbot to provide a comprehensive summary of an entire PDF or document in a single interaction, streamlining information retrieval.
**Youtube Link**

[Github Link] WebGPT: WebGPT is a powerful tool enabling users to pose questions that require internet searches. Leveraging GPT models:

  • It identifies and executes the most relevant given Python functions in response to user queries.
  • The second GPT model generates responses by combining user queries with content retrieved from the web search engine.
  • The user-friendly interface is built using Streamlit.
  • The web search supports diverse searches such as text, news, PDFs, images, videos, maps, and instant responses.
  • Overcoming knowledge-cutoff limitations, the chatbot delivers answers based on the latest internet content.
**Youtube Link**

[Github Link] WebRAGQuery: WebRAGQuery is a powerful chatbot, built with OpenAI GPT model in chainlit user interface, that harnesses the power of GPT agents, function calling, and RAG to offer an enhanced conversational experience. Here's how you can make the most of its diverse functionalities:

  • Normal ChatGPT Interaction: Engage in natural conversations as you would with a regular ChatGPT app, experiencing seamless communication.
  • Chat with Websites using RAG: Unleash the potential of the #RAG technique by sharing a website link with the chatbot. Initiate in-depth discussions and pose detailed questions about the content of the specified website.
  • Website Summarization: Request the chatbot to provide a comprehensive summary of an entire website in a single interaction. Streamline your information retrieval process with this efficient summarization feature..
  • Web Search with DocDocGo Integration: Access the vast knowledge of the web by leveraging the chatbot's connection to the DocDocGo search engine. Simply input your query, and watch as the chatbot fetches relevant links and information from the web in response.
**Youtube Link**

Time series forecasting:

[Github Link] NextX: Forecasting wind speed for the Next X Hours using different custom design and SOTA models.

[Link] Attention mechanism: Developing a custom layer for attention mechanism using Keras and seeing an example of attention mechanism in traffic forecasting.

Generative AI:

[Github Link-Tensorflow] - [Github Link-Pytorch] VAE fashion_mnist: Implementing a Variational Autoencoder (VAE) to generate fashion MNIST images.

[Github Link] GAN fashion_mnist: Training a Generative Adversarial Network (GAN) using the Fashion-MNIST dataset to generate fake images.

[Github Link] DCGAN: Training a Deep Convolutional Generative Adversarial Network (DCGAN) to generate celebrities pictures.

Natural language processing:

[Github Link] GPT2 Text Generation: Using GPT-2 to generate text based on an open-source dataset of movie summaries.

[Github Link] BERT sentiment analysis: Performing sentiment analysis on a dataset of movie reviews using the BERT (Bidirectional Encoder Representations from Transformers) language model.

[Github Link] Chatbot: Using DialoGPT_medium and DialoGPT_large pre-trained models to make chatbots. The code also applies some filters to the responses that the models will genereate. (Please note that the conversations generated by this code are not intended to make sense, as this demonstration serves purely as an illustrative example.)

[Github Link] Predict the next word: Training a language model based on LSTM architecture to generate text data, specifically movie dialogue.

Computer vision:

[Github Link] OpenCV real-time face detection: Using the Haar Cascade face detection algorithm to detect faces in each frame of the video stream from the webcam.

[Github Link] Image classification: Performing image classification on two datasets (cifar10 and mnist) using Convolutional Neural Network.

[Github Link] U-NET Image-Segmentation: Designing a U-Net to perform semantic image segmentation on the CARLA self-driving car dataset, and apply sparse categorical crossentropy for pixelwise prediction.

Optimization:

[Github Link] Optimization Algorithms: Implementation of some single and multi-objective optimization algorithms in Python (Genetic Algorithm, Evolutionary Programming, Salp Swarm Algorithm, NSGAII, etc.)

Deep Learning Infrastructure:

[Github Link] Distributed training: Distributing Tensorflow and PyTorch models over a single/multiple available GPUs.

[Github Link] Dockerized Flask ChatBot: A Flask chatbot application that utilizes the DialoGPT language model by Microsoft. The chatbot can generate responses based on user prompts and offers an interactive conversation experience.

Workflow Management:

[Github Link] MLflow experiment: MLflow-based pipeline for efficient Machine Learning experimentation and tracking.

[Github Link] Sample Luigi pipeline: Experimenting Luigi pipeline for sequential and parallel tasks.