Subscribe and Access : 5200+ FREE Videos and 21+ Subjects Like CRT, SoftSkills, JAVA, Hadoop, Microsoft .NET, Testing Tools etc..

Batch Date: Dec 16th @9:30PM

Faculty: Mrs. Sasmitha (Working as a Lead AI ML Architect with 12yrs of Exp..in AI ML)

Duration: 45 Days

Venue :
DURGA SOFTWARE SOLUTIONS,
Flat No : 202, 2nd Floor,
HUDA Maitrivanam,
Ameerpet, Hyderabad - 500038

Ph.No: +91 - 8885252627, 9246212143, 80 96 96 96 96

Syllabus:

Full Stack Agentic AI Engineering Course
Build & Deploy Production Level AI Agents
with Hands On Projects

Module 1: Agentic AI Essentials

• Agentic AI Introduction
• AI Agents vs. Agentic AI
• Comparison: Agentic AI, Generative AI, and Traditional AI
• Agentic AI Building Blocks
• Autonomous Agents
• Human‑in‑the‑Loop Systems
• Single‑Agent and Multi‑Agent Systems
• Framework overview: LangChain, LangGraph, LlamaIndex
• Ethical and Responsible AI
• Agentic AI Best Practices
• AI Implementation Success Stories: Case Studies
• Understanding Agentic AI concepts
• Identifying agent capabilities and limitations
• Choosing the right agent approach for a problem
• Responsible AI fundamentals

Module 2: Agentic AI Architectures and Design Patterns

• Architecture types: router, planner‑executor, supervisor‑worker
• Perception, cognition, action, and security modules
• Reflection / self‑critique pattern
• Tool‑use pattern
• Planning pattern
• ReAct
• Design considerations: latency, cost, reliability, auditability
• Designing secure and scalable agent architectures
• Implementing agent patterns correctly
• Building systems with predictable behavior

Module 3: Working with LangChain and LCEL (Production Pipelines)

• Data ingestion and document loaders
• Text splitting strategies
• Embeddings + vector databases
• LCEL: runnables, chains, composition
• Structured outputs (schema‑first JSON)
• Reliability: validation, retries, fallbacks
• Deployment patterns (API service)
• Building composable pipelines with LCEL
• Reliable structured output extraction
• Tool integration patterns

Module 4: Building AI Agents with LangGraph (State + Memory)

• LangGraph basics
• State schema and reducers
• Branching graphs + checkpoints
• Human‑in‑the‑loop approval gates
• Memory in graphs: short vs long term, external stores
• Deployment patterns for graph workflows
• State management for agents
• Human approval workflows
• Orchestrating reliable multi-step systems

Module 5: Agentic RAG (Enterprise)

• Agentic RAG vs traditional RAG
• Adaptive retrieval strategies
• Query rewriting + reranking
• Evidence‑first answers (citations required)
• Evaluation signals: groundedness, citation coverage, retrieval recall
• Implementing robust RAG systems
• Measuring and improving retrieval quality
• Reducing hallucinations via evidence constraints

Module 6: Knowledge Graphs for AI Systems (Neo4j)

• Entity‑relation modeling fundamentals
• Ontology‑driven design
• Provenance and audit trails in graphs
• Constraints and schema design for AI systems
• Structured reasoning using relationships
• KG modeling for enterprise AI
• Explainability via graph traces
• Governance‑ready data representations

Module 7: GraphRAG (Neo4j + Hybrid Retrieval)

• Graph traversal retrieval (multi‑hop reasoning)
• Hybrid retrieval: vector + full-text + graph traversal
• “Show evidence + show path” response pattern
• Relationship‑aware retrieval for complex questions
• Designing GraphRAG systems
• Hybrid retrieval strategies
• Explainable reasoning

Module 8: MCP (Model Context Protocol) + Tool Standardization

• MCP architecture and tool contracts
• MCP servers and standardized tool interfaces
• Secure tool access (read-only by default)
• Auditable tool execution
• Standardizing tool access for agents
• Secure tool design
• Building extensible agent tooling

Module 9: Tooling & Safety Engineering (Guardrails)

• Threat model: prompt injection + unsafe tool calls
• Input/output validation
• Allowlists and views-only execution patterns
• Human approval workflows (approve/edit/reject before execute)
• Building safe tool-using agents
• Guardrails that work in production
• Approval-driven execution patterns

Module 10: LLM Evaluation + Observability (Langfuse)

• Evaluation targets: groundedness, citation coverage, schema validity
• Judge-model patterns
• Langfuse: traces, spans, runs
• Prompt/version tracking
• Experiments and comparisons
• Cost and latency monitoring
• Evaluating agent correctness and reliability
• Instrumenting agents with Langfuse
• Running experiments safely and repeatably

Module 11: No‑Code / Low‑Code Agents (n8n)

• Building workflows with triggers and approvals
• Integrations: webhooks, email, Slack/Teams, DB calls
• Audit logs and enterprise workflow patterns
• Security considerations in automation
• Workflow automation using AI
• Integrating agents into business systems
• Approval-driven automations

Module 12: Databricks Governance + Build a Text‑to‑SQL Coding Agent

• Unity Catalog fundamentals (catalog/schema/table, permissions)
• Views-only pattern (do not query base tables directly)
• Schema-aware SQL generation (columns + datatypes injected)
• Approval gate before executing generated SQL
• Write results to Delta + run registry
• Building governed Text‑to‑SQL agents
• Schema-aware query generation
• Auditability and compliance patterns on Databricks

Portfolio Projects

Project 1 — Enterprise Knowledge Intelligence Agent (GraphRAG)

• Ingest docs → entities → Neo4j provenance graph
• Hybrid retrieval: vector + graph traversal
• Explainability: evidence + graph path

Project 2 — Multi‑Agent Operations Workflow (LangGraph)

• Supervisor→worker orchestration with retries
• Human approvals + audit logs
• Governed tool usage with allowlists

Project 3 — MCP Tool‑Augmented Agent (Tool Standardization)

• MCP server exposing safe tools (read‑only, allowlists)
• Structured tool outputs + governance patterns
• Security: least privilege + auditable execution

Project 4 — Databricks Ask‑to‑Query Agent (Genie + Warehouse)

• Build a Text‑to‑SQL coding agent (Genie-assisted)
• Inject schema (catalog.schema.table columns + datatypes) into prompts
• Validate SQL: SELECT-only + allowlisted view + LIMIT
• Approval gate: approve/edit/reject before execution
• Execute approved queries and write results to Delta + run registry

Project 5 — LLM Evaluation & Observability (Langfuse)

• Instrument traces, spans, and runs for every agent workflow
• Track prompt versions and run experiments
• Implement evaluation checks: groundedness, citation coverage, schema validity
• Monitor cost and latency; build dashboards for reliability