Generative AI 


for Software Developers

Nanodegree program syllabus

udacity.com
Generative AI Nanodegree program

Overview
In an era where generative AI is reshaping the landscape of software development, the Generative AI
Nanodegree program equips professionals with essential skills to apply these cutting-edge technologies to
enhance their software applications. The program addresses the growing demand for expertise in adapting
generative foundation models, crafting custom applications using large language models (LLMs), and
employing advanced techniques like Stable Diffusion. If you aspire to supercharge your software development
and stay at the forefront of artificial intelligence and deep learning advancements, this program is for you!

Prerequisites Length of Program*: Skill level:

A well-prepared learner: 4 months Intermediate
Uses Python at an intermediate level
Uses SQL at an intermediate level Software/Hardware and
version requirements:
Educational Objectives For this Nanodegree program, you will need access
to the Internet.

A graduate of this program will be able to: Additional software such as Python and common
Situate generative AI within the broader history, context, and deep learning libraries (e.g. PyTorch and Hugging
applications of artificial intelligence and deep learning Face) will be required, but the program includes
Udacity Workspaces with all of the relevant packages
Adapt generative foundation models to perform tasks in installed, so students will not need to download any
novel contexts additional software.
Use LLMs and prompt engineering to create a custom
chatbot
Use image generation models such as Stable Diffusion to *The length of this program is an estimation of total hours
perform image inpainting the average student may take to complete all required
coursework, including lecture and project time. If you spend
Build applications that use LLMs, implement semantic search about 5-10 hours per week working through the program, you
with vector databases, and apply retrieval augmented should finish within the time provided. Actual hours may vary.
generation techniques

uda c it y . c om
G e n e r at i v e AI N a nodegre e pro gr a m

Course #1

GenAI Fundamentals

A "crash course" in deep learning and how generative AI

compares and contrasts with previous AI techniques. Then

some hands-on practice with model adaptation.

S k i l l s c ov e r e d i n c l u d e :

What is GenAI?

Examples of GenAI
Deep Learning Fundamentals
Foundation Models
Domain Adaptation
ChatGPT & BERT

P RO J E CT # 1

Apply Lightweight Fine-Tuning to a Foundation Model

In this project, learners will explore advanced techniques in AI model fine-tuning with Hugging Face and PyTorch

tools.

To pass this project, learners will need to execute the following core tasks:

Load a foundation model

Identify and load a Hugging Face dataset for your particular task

Utilize a state-of-the-art technique to adjust the foundation model's weights to meet the needs of your task,

using a lightweight (AKA parameter-efficient) fine-tuning technique that improves performance faster and more

efficiently

udacity.com
Gener at ive AI Na node g ree p ro g r am

Course #2
LLMs & Text Generation
Dive deeper into how computers understand and create
language, and build a custom chatbot using unsupervised
machine learning and prompt engineering.

Skills covered include:

Describe business applications of LLMs
Design prompts for LLMs
Custom Datasets for LLMs
Perform prompt engineering with Python
NLP Fundamentals
Transformers & Attention Mechanisms
Retrieval Augmented Generation

PROJECT #2
Build a Custom Chat Bot
Apply retrieval-augmented generation (RAG) to create a custom OpenAI chatbot without fine-tuning.
To pass this project, learners will need to execute the following core tasks:
Find and prepare a dataset that augments a foundation model's knowledge, from a source such as APIs, web
scraping, or documents on hand
Create a semantic search pipeline by implementing a custom Python vector similarity search algorithm to
match user questions to relevant parts of the custom dataset
Compose a custom query by combining the semantic search results with the user's question and send it to the 

foundation model

u d a c i ty. c o m
Gener at ive AI Na node ree ro r
g p g am

Course #3
Computer Vision &
Generative AI
Learn how computers process and understand image data,
then harness the power of the latest Generative AI models
to create new images.

Skills covered include:

Describe key image processing techniques
Image Generation & Generative Adversarial Networks
(GANs)
Transformer-Based Computer Vision Models
Diffusion Models
Build a simple diffusion algorithm

PROJECT #3
AI Photo Editing with Inpainting
Use the Segment Anything Model (SAM) and Stable Diffusion to replace parts of images with AI-generated content.

To pass this project, learners will need to execute the following core tasks:
Create a segmentation mask by differentiating between the subject and background of an image and create a
matrix of pixels indicating the locations of these two components
Given a text prompt and the pixel locations of the subject or background, replace part of the image with an AI-
generated image
Connect your inpainting pipeline to a web interface that allows users to upload their own images and specify
their own text prompts

ud ci
a ty. co m
Generative AI N anode g ree p ro g ram

Course #4
GenAI Solutions
Learn how to design and build applications that use
generative AI models. Build applications and AI agents
using the OpenAI API, vector databases and semantic
search, the LangChain framework, and techniques for
recommendation, personalization, and retrieval
augmented generation (RAG).

Skills covered include:

OpenAI API
Prompt design
Vector databases
Semantic Search
LangChain
Retrieval Augmented Generation

PROJECT #4
Personalized Real Estate Agent
Build a “real estate agent” application that uses LLMs for content generation, vector databases, semantic search
and RAG techniques to transform standard real estate listings into personalized narratives.
To pass this project, learners will need to execute the following core tasks:
Generate synthetic data using LLMs
Embed property listing data in a vector database
Perform semantic search over property listings against user preferences
Design prompts and use RAG techniques to deliver personalized recommendations

u dac i t y . c o m
G e n e r at i v e A I N a n o d e g r e e p ro g r a m

Meet your instructors

Sergei Kozyrenko Chuyi Shang
Senior Staff Engineer Machine Learning Researcher

Giacomo Vianello Annabel Ng

Distinguished Scientist Machine Learning Researcher

Emily McMilin Derek Xu

Research Scientist Machine Learning Researcher

Brian Cruz Nathaniel Haynam

Head of AI Engineering, Advocate Machine Learning Researcher

Erick Galinkin Valerie Scarlata

AI Security Researcher Technical Content Developer

Victor Geislinger Chang She

Machine Learning Engineer CEO & Co-founder

Jason Lin
Research Engineer

udacity.com
Learn more at
w w w.udacity.com