Advanced Certificate in Generative AI

Python Programming Constructs: Conditionals, Loops, Functions

UDFs, Best Coding Practices and Exception Handling

Python for Data Science and Pandas: Working with relational databases, Data Cleaning, Preprocessing, Analysis

Advanced Text Processing using Pandas

Basics of Linux: Commands, Setting up Local Environment

Apply prompting techniques to create prompts for asking questions and evaluating the customer's response

Establish metric(s) to measure model performance

Prompt Engineering: Improve the assistant's responses by applying simple (non-reasoning) prompting techniques

Prompt Engineering: Improve the assistant's accuracy by applying Chain of Thought reasoning-based prompting techniques

Apply fine-tuning using OpenAI APIs to train an LLM on your custom data

Learn the best practices for fine tuning OpenAI APIs

Transfer learning: Apply the same principles to other problems in your domain

Deploy and launch ShopAssistAI application on Flask

Iterate and improve the UI of the app using ChatGPT's code writing capabilities

Create and analyse embeddings for semantic search

Understand the entire semantic search pipeline including chunking, embedding, and retrieval

Create embeddings for large documents by creating chunks

Create a Q/A system that fetches answer using similarilty search over embeddings

Scale the Q/A system by making use of vectorstores like ChromaDB

Embed, index large documents and search in Vectorstore

Integrate LLM chat models like GPT with the semantic search to build a retrieval augmented generation system that directly responds to user queries

Experiment with different vectorstores, search and index algorithms and LLMs to improve the chatbot

Understand the different parts of LangChain - Models, Prompts, Indexes, Chains, Memory and Agents

Explore the different tools in LangChain and initialise an agent that uses the tools to read different types of files or data present in the company database

Build the backend for the system using Vectorstore options present in LangChain

Divide the documents into chunks and apply the LLM to create the embeddings and extract entity for the chunks of document and store them in the Vectorstore

Construct the Search Index and Entity Store and create a functionality to update it with every question that the user asks

Use the Chain functionality of LangChain to connect all the components

Evaluate the results and improve them by experimenting with different LLMs, indexing and embedding algorithms

Explore other agents and tools to improve the system like adding features like automatic email notifications on some issues, etc.

Identify the cloud services required for creating the scalable system

Expose the system through a chat based front end to the user

Multimodal Learning: Audio, Image, Text, Heatmap among others within a LLM

Understand and implement legacy image generation models such as variational autoencoders and generative adversarial networks

Understand the components of diffusion models and the process by which images are generated and work on building a stable diffusion pipeline component-by-component

Set up a simple stable diffusion pipeline and create suitable prompts for image generation and use the model to generate relevant images

Use vector embeddings to solve a real-world use-case problem based on semantic similarity

Fine-tune language generation models for a particular problem statement and evaluate the model