Postgraduate Certificate in Generative AI and Agentic AI

Module 5: Machine Learning and Neural Network

• Introduction to machine learning

• Introduction to neural network

• Understand concepts of NN: Layers, tokens, embeddings, gradients, back propagation

Module 6: NLP and Transformer Architecture

• How the original Transformer architecture works

• Identify appropriate use cases and limitations of GenAI systems

• Understand different LLM architectures

• Understand key components of transformer model

Module 7: Training and Reasoning Models

• Understand components of transformer

• In-depth code analysis of transformer

• Pre-training, Post-training of LLM models

• How reasoning models work

Module 8: Pydantic and Langchain

• Define data models with Pydantic Base Model

• Validate and handle type errors automatically

• Call LLMs through LangChain's chat interface

• Build reusable prompt templates with variables

• Parse LLM responses into Pydantic models

Module 9: GenAI Tools, Prompt Engineering & Frameworks

• Apply prompt engineering techniques (zero-shot, few-shot, chain-of-thought, ReAct) to improve LLM output quality

• Integrate LangChain and PydanticAI to structure and chain LLM interactions

• Evaluate prompt effectiveness through systematic comparison

• Understand structured output generation and validation

Module 10: Vector Database, Retrieval Augmented Generation and Tool calls

• Understand vector databases

• Implement storage and indexing strategies for embedding vectors

• Understand different chunking techniques and its effect

• Integrate tool calling/function calling to extend LLM capabilities with external data sources

Module 11: Eval & Synthetic Data Generation

• Establish evaluation frameworks using RAGAS metrics

• Implement performance monitoring using LangSmith or equivalent observability tools

• Generate synthetic evaluation datasets that represent realistic query patterns

• Analyze RAG system performance bottlenecks and implement optimization strategies

Module 12: Context Engineering & LLM Monitoring

• Differentiate between sparse (BM25, TF-IDF) and dense (embedding-based) retrieval approaches

• Implement hybrid retrieval systems that combine sparse and dense methods for improved accuracy

• Build multi-hop retrieval systems that answer complex queries requiring multiple information sources

• Apply RAPTOR (Recursive Abstractive Processing for Tree-Organized Retrieval) for hierarchical document processing

• Design memory systems incorporating both long-term (vector DB) and short-term (conversation history) context

Module 13: Reasoning, caching and guardrails

• Configure prompt caching to reduce latency and API costs for repeated context

• Implement guardrails for content filtering, safety controls, and output validation

• Optimize inference costs through strategic use of different model tiers

• Evaluate trade-offs between model capabilities, speed, and cost for production use cases

Module 14: Building & Deploying Knowledge-based Applications

• Build production-ready RAG applications using Chainlit, Streamlit, or Next.js frameworks

• Design user interfaces that effectively display retrieved context and generated responses

• Implement authentication, rate limiting, and usage tracking for multi-user deployments

• Structure knowledge bases with metadata, versioning, and update mechanisms

Module 15: Model fine-tuning

• Distinguish between when to use RAG, prompt engineering, agents or fine-tuning for specific use cases

• Prepare and curate high-quality datasets for domain-specific model fine-tuning

• Configure and execute fine-tuning jobs using Arcee's platform for specialized model adaptation

Module 16: Generative AI in Business & Society

• Analyze industry-specific GenAI use cases

• Evaluate business value and ROI of GenAI solutions

• Identify organizational and adoption challenges

Module 17: Tools & Frameworks for Agentic AI

• Explain the core concepts of agentic AI: autonomy, tool use, planning, and reasoning loops

• Compare agentic frameworks (LangGraph, CrewAI, OpenAI Assistants API)

• Build a simple agent with tool-calling capabilities using LangGraph

• Understand when agentic solutions are appropriate vs. RAG or fine-tuning

Module 18: Agent Architectures

• Compare autonomous agents (goal-driven, self-directed) versus workflow agents (predefined paths)

• Evaluate risk-reward trade-offs between agent autonomy levels

• Create hybrid systems that combine autonomous reasoning with workflow safety nets

• Create agent skill libraries organized by domain (web interaction, data processing, communication)

Module 19: Multi-Agent Systems

• Architect multi-agent applications with specialized agent roles

• Design coordination mechanisms (supervisor, hierarchical, democratic patterns)

• Design scratchpad/working memory systems for multi-step reasoning

• Configure summary memory to compress long conversations efficiently

• Implement MCP servers and clients with proprietary data sources

Module 20: Deep research and agent evals

• Build multi-agent research systems that synthesize information from multiple sources

• Configure A2A servers for standardized inter-agent communication

• Build authentication and authorization for secure agent interactions

• Implement agent cards (metadata about agent capabilities and interfaces)

Module 21: Multimodal AI

• Understand how multimodal LLMs process images, audio, and video alongside text

• Implement vision-language pipelines for image understanding, OCR, and visual Q&A

• Apply multimodal prompting techniques to extract structured data from documents and images

• Evaluate multimodal model capabilities, limitations, and appropriate use cases

Module 22: Case Studies & Mini-Project

• Apply agentic AI concepts to real-world problem scenarios

• Test and debug agent systems for reliability

• Prepare a functional prototype for capstone readiness

Module 23: Responsible & Ethical AI

• Identify ethical risks and biases in AI systems

• Apply responsible AI principles to system design

• Evaluate AI deployments for compliance and governance

• Design safety guardrails for autonomous agent behavior

Module 24: Capstone Project – Build & Iterate

• Design an end-to-end Agentic AI solution architecture

• Implement automated workflows using agents and tools

• Setup evaluation metrics

• Test, iterate, and refine agent behavior based on feedback

Module 25: Capstone Project – Final Submission

• Finalize and deploy a complete Agentic AI solution

• Document architecture, design decisions, and outcomes

• Submit and present the completed capstone project