Scribd
Upload
21 views9 pages

Master Catalog For GenAI Programs For LNW-19Jul2024

The document outlines a comprehensive course on Machine Learning basics essential for Generative AI, covering topics such as Foundation Models, working with Large Language Models (LLMs), AI-assisted software development, and recent advancements in the field. It includes subtopics like supervised and unsupervised learning, prompt engineering, AI infrastructure, and various applications in industries like media and entertainment. The course also features hands-on projects and emphasizes architecture and design considerations in AI applications.

Uploaded by

manohara.devara
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as XLSX, PDF, TXT or read online on Scribd
21 views9 pages

Master Catalog For GenAI Programs For LNW-19Jul2024

The document outlines a comprehensive course on Machine Learning basics essential for Generative AI, covering topics such as Foundation Models, working with Large Language Models (LLMs), AI-assisted software development, and recent advancements in the field. It includes subtopics like supervised and unsupervised learning, prompt engineering, AI infrastructure, and various applications in industries like media and entertainment. The course also features hands-on projects and emphasizes architecture and design considerations in AI applications.

Uploaded by

manohara.devara
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as XLSX, PDF, TXT or read online on Scribd
You are on page 1/ 9

Topic

Machine Learning Basics required for Generative AI

Generative / Modern AI with Foundation Models


(FMs)

Working with LLMs

AI assisted Software Development/ SDLC

Recent Advancements & Use cases

Note:
Sub Topics
General, Narrow AI and Generative AI
Basics of Machine learning- Supervised, Unsupervised, Reinforcement Learning
Deep Neural Networks & Transformers - What Business needs to understand
Applications of Machine Learning in Media & Entertainment, Gaming

Basics of Foundation Models


How Foundation Models is disrupting Tech/IT & Industries and society/ careers
Multimodal Large Language Models
Proprietary LLMs & Open source LLMs
Tokenization, Embeddings in LLMs
Reinforcement learning from Human Feedback
Deploying and scaling LLMs
Where to use Foundation Models- Business use case/ value realization categories

Zero shot, one shot, few shot prompting/ Prompt Engineering


BreakOut Room Activity: Prompt Engineering/ Testing Proprietary and open LLM capabilities with reasoning, m
Finetuning-Full Parameter and Parameter Efficient and its application to your business
Retrieval Augmented Generation with Vector databases and its business applications
Chatbots, AI Agents with LLMs & AI agentic tools
AI Copilots
AI Infrastructure for Generative AI/ LLMs
LLMOps/ Deployment & Scaling
Walkthroughs- AI services on Azure, GCP, Databricks (Bedrock, Vertex AI, Azure AI Studio)
LLM Cost/ Pricing

Code generation, unit test case development, documentation, debugging using AI Coding assistants
AI assisted testing using LLMs
AI assisted CI/CD - DevOps
AI assisted Observability, Networking & Security

AI Agentic Workflows- Reflection, Tool Use, Planning, Multi-agent Collaboration


Digital Humans and Humanoid Robots - Transforming Customer service & other use cases
Use cases of Generative AI/ Foundation Models -Media & Entertainment, Gaming

Total Duration
1 hr of Q&A is included in the above schedule.
Duration

13
Course Topic

Features of open source LLMs & Proprietary LLMs

Subword Tokenizers and Embedding models with demos

Vector database design patterns

Designing/ Architecting Finetuning strategies (incl Huggingface libraries)

Finetuning with parameter efficient methods- Concept with Demo

Designing/ Architecting production grade Retrieval Augmented Generation (RAG) pipelines on AWS,
Azure, GCP, Databricks

Designing/ Architecting LLM based apps with Langchain & Llamaindex

Performance benchmarks/ evaluation of LLMs

LLM Security & Governance- What an AI Architect needs to know for designing secure solutions

Post production observability and model optimization (incl handling model drift, DLP)

AI Agents with Langgraph, Autogen, llama-agents, AI Agentic workflows- Reflection, Tool use
(function calling), Planning, Multi-agent collaboration
AI Agents with Langgraph, Autogen, llama-agents, AI Agentic workflows- Reflection, Tool use
(function calling), Planning, Multi-agent collaboration

Total Duration (Theory + Demos, Exercises)


Key Considerations
Coding hands-on/ demo of code will be restricted to 15% of the total time of the program. More emphasis will be on architect
be used will usually be related to architecture & design.
There will be case solving/ case study based breakout sessions/ design challenges for participants in several of the sessions

Good performance on Accuracy, precision &Out of Scope


other model evaluation metrics is out of scope. The demos & hands-on with data
getting optimal results
Penetration testing for Security assessment
The topic durations against each topic is the planned duration. Actual duration might be different, however overall program w
Scope of LLMs to be used is as per the TOC
Duration
Sub Topics (hrs)
Working with LLM APIs
Working with open source LLMs 1
Model Selection/Eval/Deployment
Why LLMs use subword tokenizers
Context based embedding models used with LLMs 2
Counting tokens using tiktoken (Byte pair encoding)
How to work with Vector databases for AI app dev
AI App architectures with Vector databases
4
Vector database selection- pure play vs integrated/ multi modal
Building AI app with Singlestore/ MongoDB/ Milvus
Transfer Learning & full parameter finetuning
Parameter Efficient Finetuning- Basics
Catastrophic Forgetting, Quantization
2
GPU Memory calculation - loading & funetuning LLM
Multi-GPU finetuning, model parallelism, FSDP, Deepspeed
Real world examples of Finetuning use cases
Executing PEFT based finetuning with QLORA, LORA
Huggingface and other libraries for PEFT
2
Architecture & Demo - Finetuning Proprietary LLMs (OpenAI, Google Gemini)
Architecture & Demo- Finetuning Open source LLMs (Llama 3, Mistral/ Gemma)
RAG Architectures on AWS/ Azure/ GCP/ Databricks- Advanced, Modular
RAG vs Finetuning- when to use what
6
Architectural approach - Build an ArXiv RAG search system w FAISS
Using GPT-4o for parsing and retrieval for financial report RAG

Intro to Langchain and AI App architectures with AWS/ GCP/ Azure/ Databricks

6
Intro to Llamaindex and AI App architectures with AWS/ GCP/ Azure/ Databricks
When to use- Langchain vs LlamaIndex
Parsing 10k / Annual Report with LlamaParse & gpt-4o
MMLU, HumanEval, TrueQA, other techniques for LLM evaluation
2
Metrics for LLM evaluation - Perplexity, etc
Prompt Injection, content filtering, abuse monitoring, red teaming 2

Implementing rule/ ML based guardrails to LLM prompts & completions on AWS


LLM Observability & tooling for it
2
Managing LLM drift & Data loss prevention
Intro to Langgraph, Autogen, Llama-agents

3
AI Agentic Workflows
Enterprise use cases of AI Agents 3
AI Agent Architectures- AWS/ Azure/ GCP/ Databricks
Demo- Build AI Agent to query SQL databases in natural language
32 hrs
m. More emphasis will be on architecture & design related discussions and any code that will

ticipants in several of the sessions

pe. The demos & hands-on with data will be mostly to show the How-tos and not focussed on

different, however overall program will be of 32 hrs


Project 1 (No code) Build the architecture of a Conversational AI Platform (for building chat and voice bots) using Multimoda
Project 2 Design 3 use cases in your existing apps/ domain (incl 1 net new use case) and document the scope, bus
Project 3 Build a web research assistant using Langchain/ Llamaindex
Project 4 Build a customer assistance chatbot using a LLM and Langchain
Project 5 Build a Document Information Extraction app using LLMs

Choice of projects/ allocation of a project to each participant will be done during the program. Submissions to be done after p
or building chat and voice bots) using Multimodal LLM based approach. Reference- Kore.ai (Use ppt format)
net new use case) and document the scope, business problem addressed, solutioning approach in ppt format

ing the program. Submissions to be done after program ends and the submissions will be evaluated.

You might also like