### Course Overview
- **Course Title:** Practical Agentic AI: RAG, Planning & Vector Search
- **Instructor:** Shreejit Gangadharan (Learnsector LLP)
- **Target Audience:**
- Developers with Python experience
- AI engineers interested in **Agentic AI systems**
- Professionals working with **LLMs, RAG, or AI workflows**
- Intermediate to advanced learners in **AI/ML and automation**
- **Prerequisites:**
- Working knowledge of **Python** and **virtual environments**
- Comfort with **CLI & Git**
- Understanding of **HTTP APIs, JSON, and LLM prompts**
### Curriculum Highlights
- **Key Topics Covered:**
- **Agentic AI fundamentals** vs. traditional LLM chat apps
- **Perceive→Reason→Act loop** for multi-step task execution
- **Model–Controller–Prompter (MCP) workflow** orchestration
- **Retrieval-Augmented Generation (RAG)** with grounding and citations
- **Web search integration** (Tavily) and **LLM-based ranking**
- **Memory persistence** using **SQLite** and migration to **ChromaDB**
- **Topic "pillars," weighted selection, and semantic re-ranking**
- **Explainable AI (XAI)** for user-facing recommendation explanations
- **Agent risk mitigation** (prompt injection, memory poisoning, spoofing)
- **Offline testing and scenario-based evaluation**
- **Framework comparisons**: **LangChain, LlamaIndex, CrewAI, AutoGen**
- **CLI agent deployment** with reproducible configs and prompts
- **Key Skills Learned:**
- Designing **autonomous AI agents** with goal-directed workflows
- Implementing **RAG pipelines** for factual, grounded responses
- Integrating **live search (Tavily)** and **LLM ranking (ChatGPT/Gemini)**
- Building **persistent memory systems** for agent interactions
- Engineering **semantic search** with **vector embeddings** and **re-ranking**
- Developing **explainable recommendations** with user-friendly outputs
- Hardening agents against **security risks and adversarial attacks**
- Structuring **agentic projects** with modular, testable code
- Deploying **CLI-based AI tools** for real-world use cases
### Course Format
- **Duration:** 2 hours on-demand video
- **Format:** Self-paced online course with hands-on labs
- **Resources:**
- 2 downloadable resources (code templates, checklists)
- Access on **mobile and TV**
- **Certificate of completion**
### Capstone Project
- **Project Title:** CLI **"Personalized News Curator" Agent**
- **Features:**
- Preference learning from user interactions
- **Topic-pillar-based ranking** with weighted exploration
- **Semantic re-ranking** for diverse recommendations
- **Live web search** (Tavily) + **LLM summarization**
- **SQLite → ChromaDB migration** for embeddings
- **Explainable outputs** ("why this was recommended")
- **Recency decay** and **source allow-listing**
- **Autonomous loop** for continuous updates
### Tools & Frameworks
- **Primary Tools:** **LangChain, Tavily, ChatGPT/Gemini, SQLite, ChromaDB, pytest**
- **Secondary Frameworks:** **LlamaIndex, CrewAI, AutoGen**
### Instructor Background
- **Experience:** 12 years in **Flipkart, Microsoft, Google**
- **Udemy Stats:**
- **4.4 Instructor Rating** (22,277 reviews)
- **475,337 students** across 36 courses
