AI & LLM Engineering for .NET Architects

Planner & Orchestration: Automating complex multi-step AI tasks

1 Views Updated 5/4/2026

AI Planning & Autonomy

What if you ask the AI to "Send a summary of my last 5 emails to my boss"? This requires multiple steps. A Planner is an AI model that looks at your prompt, looks at your available plugins, and writes its own "To-Do" list.

1. Function Calling Stepwise Planner

This is the most modern planner in SK. It acts like a loop:

  1. Thought: "I need to fetch the emails first." (Call Plugin A)
  2. Observation: "Got 5 emails."
  3. Thought: "Now I need to summarize them." (Call Plugin B)
  4. Thought: "Now I need the boss's email address." (Call Plugin C)
  5. Final Action: "Send email."

2. The Risk of Self-Orchestration

Planners are powerful but dangerous. They can get stuck in infinite loops or hallucinate function calls. As an architect, you must set **MaxIterations** and provide a very limited, secure set of tools to the planner.

4. Interview Mastery

Q: "What is the difference between a 'Static' and 'Dynamic' Planner?"

Architect Answer: "A **Static Planner** (like the old Sequential Planner) creates the whole plan upfront and then executes it. It's fast but brittle—if one step fails, the whole plan is ruined. A **Dynamic Planner** (like Stepwise) decides the next step ONLY after seeing the result of the previous one. It is much smarter and more resilient to errors, but it uses more tokens and takes longer to finish."

AI & LLM Engineering for .NET Architects
1. AI Foundations & Prompt Engineering
The LLM Landscape: Transformers, Attention, and Tokens Advanced Prompt Engineering: Few-shot, Chain-of-Thought, and ReAct Prompt Versioning & Management in Production LLM Cost Estimation: Token accounting and budget strategies
2. Semantic Kernel & Integration
Introduction to Microsoft Semantic Kernel (SK) Skills & Plugins: Extending the LLM with native C# functions Planner & Orchestration: Automating complex multi-step AI tasks Connectors: Switching between OpenAI, Azure OpenAI, and HuggingFace
3. Vector Databases & RAG
The RAG Pattern: Solving the 'Static Knowledge' problem Embeddings Deep Dive: Converting text to math Vector DBs: Azure AI Search vs Pinecode vs Milvus Hybrid Search: Combining Keyword and Semantic search for accuracy
4. Advanced RAG Techniques
Document Chunking Strategies: Overlap, Slidewindow, and Semantic splitting Recursive Document Processing for massive knowledge bases Context Window Management: Summarization vs Truncation Citations & Grounding: Ensuring the AI doesn't hallucinate
5. AI Safety & Guardrails
Content Moderation: Azure AI Content Safety integration Prompt Injection: Defending against adversarial attacks Punitiveness & Bias: Evaluating and mitigating model behavior Self-Correction Patterns: Letting the AI check its own work
6. Small Language Models (SLMs) & Local AI
The rise of SLMs: Phi-3, Llama-3-8B, and Mistral Running AI Locally with ONNX and LocalLLM Quantization: Running 70B models on 16GB RAM Edge AI: Deploying models to local devices and private clouds
7. Multimodal & Agentic AI
Multimodal AI: Processing Images, PDFs, and Audio in C# Agentic Workflows: Multi-agent collaboration with AutoGen Function Calling: Letting the LLM use your SQL and API tools Memory Management: Ephemeral vs Long-term Semantic memory
8. FAANG AI Engineer Interview
Case Study: Designing a Global Enterprise AI Knowledge Assistant Case Study: Building an Autonomous AI Agent for Software Dev