Skip to main content
Back to Programs
⚙️Flagship

AI Engineer (Agentic)

Full-Stack Production AI Engineering

The most comprehensive AI engineering program available. You will go beyond API calls and build the systems that power real-world AI products: scalable RAG pipelines, knowledge graphs, autonomous multi-agent systems, fine-tuned domain-specific models, and the full MLOps infrastructure to operate them in production. Every week pairs deep technical theory with a hands-on production-grade lab. By graduation you will own the full AI engineering lifecycle.

Duration

6 months

Level

Intermediate → Advanced

Starts

September 1, 2026

Modules

6 modules

Chat on WhatsApp

What You'll Learn

Transformer internals, attention, KV-cache, quantization (GPTQ/AWQ/GGUF)
Production RAG: hybrid search, reranking, RAPTOR, Self-RAG, Agentic RAG
Knowledge Graphs: Neo4j, Cypher, Graph-RAG, multi-hop reasoning
Agent architectures: ReAct, Plan-and-Execute, multi-agent systems
Fine-tuning: LoRA, QLoRA, DPO, RLHF, synthetic data generation
Evaluation: RAGAS, LLM-as-Judge, HELM, red-teaming
Production MLOps: vLLM, TGI, Triton, observability, cost optimization

Full Curriculum

Detailed Syllabus

01
Weeks 1–4

LLM Foundations & Inference Engineering

  • Transformer architecture: multi-head attention, FFN, layer norms, positional encoding
  • Tokenization deep-dive: BPE, WordPiece, SentencePiece, tiktoken
  • Sampling strategies: temperature, top-p, top-k, beam search, speculative decoding
  • KV-cache: mechanics, memory budgeting, cache eviction strategies
  • Quantization: GPTQ, AWQ, GGUF — trade-offs in latency vs quality
  • Prompt engineering: system prompts, few-shot, chain-of-thought, tree-of-thought
  • Structured outputs: JSON mode, function calling, tool use, constrained generation
  • Guardrails & safety: input/output filtering, constitutional AI principles
  • Lab: Deploy and benchmark a quantized LLM (Llama 3 / Mistral) on GPU; latency profiling
02
Weeks 5–9

Production RAG Systems

  • Document parsing: PDF, HTML, tables — layout-aware extraction (Unstructured, LlamaParse)
  • Chunking strategies: fixed, recursive, semantic, sentence-window, late chunking
  • Embedding models: OpenAI text-embedding-3, Cohere embed-v3, BGE-M3, E5-large
  • Vector databases in depth: Qdrant, Pinecone, Weaviate, pgvector — selection criteria
  • Hybrid search: dense + sparse (BM25, SPLADE) with Reciprocal Rank Fusion
  • Cross-encoder reranking: Cohere Rerank, ColBERT, BGE-reranker
  • Query transformations: HyDE, multi-query, step-back prompting, query decomposition
  • Advanced RAG: RAPTOR (tree-level summarization), Self-RAG, Corrective RAG, Agentic RAG
  • RAG evaluation: RAGAS framework — faithfulness, answer relevance, context recall
  • Lab: Build a production multi-tenant RAG system with hybrid search, reranking, and RAGAS eval
03
Weeks 10–13

Knowledge Graphs & Graph-RAG

  • Graph theory fundamentals: nodes, edges, properties, traversal algorithms
  • Ontology design: classes, properties, constraints, OWL/RDFS patterns
  • Neo4j and Cypher: CRUD, pattern matching, path queries, aggregations
  • Entity extraction with LLMs: NER, coreference resolution, relation mining
  • Knowledge graph construction pipelines: ingestion → extraction → validation → storage
  • Graph-RAG: retrieval over knowledge graphs, Cypher generation from natural language
  • Multi-hop reasoning: path-based answers, subgraph retrieval
  • Combining vector search with graph traversal (hybrid retrieval)
  • Lab: Build a Graph-RAG system: ingest docs → extract entities → Neo4j → LLM Q&A
04
Weeks 14–18

AI Agents & Multi-Agent Systems

  • Agent architectures: ReAct, Plan-and-Execute, Reflexion, LATS
  • Tool use: function calling, API tools, code execution, web browsing
  • Memory systems: in-context buffers, external vector stores, episodic memory
  • LangGraph: stateful agent graphs, conditional edges, persistence, human-in-the-loop
  • Multi-agent systems: hierarchical orchestration, collaborative and competitive patterns
  • Task decomposition, delegation, and consensus mechanisms
  • Agent evaluation: trajectory eval, tool-use accuracy, task completion metrics
  • Safety: sandboxing, permission scoping, cost caps, rate limiting, circuit breakers
  • Lab: Build a multi-agent research system with web tools, memory, and LangGraph orchestration
05
Weeks 19–22

Fine-Tuning & Alignment

  • When to fine-tune: fine-tune vs. RAG vs. prompt engineering decision framework
  • Full fine-tuning vs. PEFT: LoRA, QLoRA, IA3, prefix tuning
  • Dataset curation: quality filtering, deduplication, format conversion (Alpaca, ShareGPT)
  • Synthetic data generation with LLMs: self-instruct, Evol-Instruct, Magpie
  • Supervised fine-tuning with Hugging Face Trainer and TRL SFTTrainer
  • RLHF pipeline: reward modeling, PPO, GRPO
  • DPO and preference optimization: SimPO, KTO, ORPO
  • Evaluation: HELM, lm-eval-harness, MT-Bench, LLM-as-Judge
  • Red-teaming: adversarial prompts, jailbreaks, safety benchmarks
  • Lab: Fine-tune Llama 3 with QLoRA on domain data; evaluate with lm-eval
06
Weeks 23–26

Production MLOps & Capstone

  • Inference servers: vLLM (PagedAttention, continuous batching), TGI, Triton Inference Server
  • Serving architecture: router, load balancer, autoscaler, shadow deployment
  • Cost optimization: quantization, batching, prompt caching, tiered routing
  • Observability: OpenTelemetry tracing, Langfuse, Arize Phoenix, custom dashboards
  • CI/CD for LLMs: test-driven prompt development, regression testing, canary rollouts
  • LLM gateway: rate limiting, key management, usage tracking, fallback routing
  • Capstone: Design and deploy a full agentic AI system — RAG + agents + knowledge graph + fine-tuned model + MLOps — production-grade on GCP

Outcomes

After this program, you'll be able to:

  • Build and deploy production LLM inference stacks (vLLM, TGI) on GPU infrastructure
  • Architect and implement multi-stage RAG pipelines with hybrid search and evaluation
  • Build Graph-RAG systems combining Neo4j knowledge graphs with vector retrieval
  • Design and deploy multi-agent systems using LangGraph with memory and tool use
  • Fine-tune open-source LLMs (Llama 3, Mistral) using QLoRA/DPO on domain data
  • Implement production MLOps: CI/CD, observability, cost optimization, canary deployments
  • Own the complete AI engineering lifecycle from prototype to production

Prerequisites

Before you start, you should have:

  • Python proficiency (functions, classes, async)
  • Basic ML/DL understanding (neural networks, backprop)
  • Familiarity with REST APIs and cloud basics (any provider)
  • Recommended: completed AI Engineering Foundations or equivalent

Ready to master AI Engineer (Agentic)?

Download the full syllabus or chat with us on WhatsApp.

WhatsApp Us