Agent Engineering Bootcamp: Developers Edition

Design and deploy intelligent retrieval systems that reason, route queries, and adapt across multi-turn conversations—far beyond naive RAG

Framework: Implement stateful RAG architectures with routers, reflection loops, and multi-hop reasoning strategies

Hands-on: Build custom routing logic that knows when cosine similarity fails and switches between retrieval strategies

Host, quantize, and serve production LLMs locally and in the cloud with cost-efficiency and low latency

Technical deep-dive: Master quantization techniques (GPTQ, GGUF, QLoRA) to reduce model size by 4-8x without sacrificing performance

Infrastructure: Deploy models using Ollama (local), RunPod (cloud), and FastAPI with auto-scaling capabilities

Outcome: Build intelligent caching layers that recognize similar queries, avoid redundant LLM calls, and improve over time.

From scratch: Code semantic distance functions using vector embeddings and proximity thresholds

Architecture: Discussion on cache hit/miss systems with reranking and feedback loops that learn from usage patterns

Outcome: Move beyond flat retrieval to structured reasoning by building graph-based memory with natural language to Cypher query generation

Modeling: Design graph schemas for agent memory, extracting entities and relationships from unstructured text

Tools: Implement with Neo4j or Memgraph and connect to RAG pipelines for context-aware graph traversal

Outcome: Deploy coordinated multi-agent workflows with agent-to-agent communication, human-in-the-loop patterns, and production guardrails.

ReAct paradigm: Build modular Reason+Act pipelines with tool use, planning, and reflection in both Python and n8n

ADK & MCP: Combine Google's Agent Development Kit with Modular Cognitive Planning for enterprise-grade orchestration