Skip to content
CMKL UniversityProfessional
Training Center

Advanced · Agentic AI & Automation

A1Agentic AI Systems and Optimization

Design, optimize, and ship secure, efficient AI agents, from architecture to production.

Who it's for

Technical professionals with a solid foundation in machine learning and deep learning, and basic familiarity with large language models (LLMs).

You'll walk away with

A working agentic AI prototype / pipeline

Design, implement, and deploy secure and efficient agentic AI systems, from architecture and memory to LLM optimization and production deployment.

Format

Lectures + hands-on workshops

Assessment

Pre- and post-tests (30 multiple-choice questions); hands-on workshops aligned to each learning outcome

Tools & platforms

LangChainLangGraphHugging FaceLangSmithvLLMDockerModel Context Protocol (MCP)

Programme Curriculum

What you'll be able to do

  • Understand the foundations of agentic AI systems, including architectures, reasoning strategies, and monitoring principles
  • Apply secure prompt engineering practices to ensure robustness, reliability, and safety in AI systems
  • Implement and deploy memory-augmented AI systems using Retrieval-Augmented Generation (RAG) and tool integration
  • Apply optimization techniques for LLMs to improve efficiency and scalability in deployment contexts

Your learning journey

  1. Module 1

    Introduction to Agentic AI and Monitoring

    • Agentic AI ArchitecturesAgentic AI; autonomy; environment; reactive, deliberative, and hybrid agents; LLM backbone; reasoning; memory; tool integration; LangChain; LangGraph; Hugging Face
    • Optimization, Observability & MonitoringModel drift; hallucination; efficiency; reliability; scalability; logging; tracing; metrics; observability; LangSmith
    • Reasoning & Secure Prompt EngineeringChain-of-Thought (CoT); ReAct; Self-Consistency; secure prompt engineering; threat awareness; agent logging and monitoring
    • Use-Case ReflectionAgentic AI use cases; domain application; memory and tool integration considerations

  2. Module 2

    Memory, Knowledge, and Tool Use

    • Retrieval-Augmented Generation (RAG)Agent memory; Retrieval-Augmented Generation (RAG); embeddings; vector databases; indexing and retrieval
    • Tool Use & Function CallingTool calling; function calling; LangGraph; tool-using agents
    • Secure RAG & Data AccessSecure knowledge integration; data leakage; unsafe retrieval; source verification; document chunking; embeddings; vector database; retrieval pipeline; basic RAG implementation
    • Use-Case ReflectionDomain-specific tool requirements and integration considerations

  3. Module 3

    LLM Optimization

    • LLM Optimization & Performance BottlenecksQuantization; Parameter-Efficient Fine-Tuning (PEFT); LoRA; sparsity; knowledge distillation; Transformer fundamentals; efficient attention
    • Quantization & PEFTLow-bit quantization; MXFP4; GGUF; LoRA; QLoRA; model quantization; parameter-efficient fine-tuning
    • Knowledge Distillation & Inference EfficiencyKnowledge distillation; student–teacher models; model compression; KV Cache; Paged Attention; long-context inference; performance profiling
    • Optimization Trade-OffsEfficiency, quality, latency, and scalability trade-offs

  4. Module 4

    Deploying Agentic AI Systems

    • AI Deployment FundamentalsSingle and batch inference; AI microservices; inference engines; vLLM; deployment configuration
    • Production ReadinessProduction scaling; monitoring; logging; Docker; containerization; safety guardrails; guard models
    • AI Safety, Guardrails & MonitoringSecure tool access; safety guardrails; monitoring and observability; human oversight
    • Scalable Model ServingModel parallelism; multi-GPU and multi-node serving; high-throughput inference; low-latency architecture; Model Context Protocol (MCP); multi-agent frameworks; end-to-end deployment
    • Agent Integration & OrchestrationAPIs; enterprise tools; tool calling; Model Context Protocol (MCP); multi-agent workflows and orchestration
    • Scaling & Next StepsProduction scaling challenges; deployment considerations; next-step planning

Remarks

  • Prerequisite: Completion of a pre-test before the first session