Building Smarter AI Agents: Context Engineering, Memory, Orchestration & Architecture Patterns #361
Description
Talk title
Building Smarter AI Agents: Context Engineering, Memory, Orchestration & Architecture Patterns
Short talk description
This workshop dives into the practical side of building AI agents that actually work in the real world. We’ll move beyond YouTube tutorials and simple demos to explore how modern AI agents are designed, orchestrated, and deployed at scale. Attendees will learn key concepts like context engineering, multi-agent collaboration, orchestration patterns, and memory management. Through hands-on examples using tools like LangGraph, CrewAI, and OpenAI’s Agent Builder, participants will gain the skills to move from prototype to production-ready AI systems — with a clear view of what works, what doesn’t, and what’s next in the agentic AI era.
Long talk description
This workshop is designed for developers, AI enthusiasts, and builders who want to go beyond surface-level “AI agent” demos and truly understand how to design, build, and deploy intelligent systems that work in the real world.
We’ll start with the fundamentals — what AI agents really are, the different types, and how they’re being used in enterprises today. From there, we’ll unpack the differences between LLMs, RAG systems, and agentic AI, and understand how agents bring reasoning, memory, and autonomy into applications.
The session will then move from prompt engineering to context engineering — the shift that’s defining the next generation of AI systems. You’ll learn how to manage context, offload reasoning, and handle the “context pollution” that often breaks agent logic.
We’ll explore single-agent and multi-agent architectures, diving into practical design patterns like ReAct, Reflection, Reflexion, and Routing, and orchestration strategies such as Sequential, Hierarchical, Consensus, and Map-Reduce. Each pattern will be explained with clear logic and real implementation examples.
On the hands-on side, we’ll explore code-based frameworks such as LangGraph, CrewAI, AutoGen, and Agents SDK, along with no-code options like n8n and OpenAI’s Agent Builder — so participants can see both developer-friendly and accessible ways to build agent systems.
We’ll also cover Model Context Protocol (MCP) — a key emerging standard for context exchange between agents — and look at how to evaluate, monitor, and govern AI agents in production.
This isn’t a theoretical or YouTube-style tutorial. It’s based on real experience from building and deploying AI agents in live environments — sharing what works, what fails, and how to make these systems reliable at scale.
By the end of the workshop, participants will have a clear roadmap of how to move from idea to production, understand the architectural patterns that matter, and gain the practical confidence to build smarter, more autonomous AI systems that deliver real value.
What format do you have in mind?
Workshop (45-60 minutes, hands-on)
Talk outline / Agenda
- Agent 101 – Understanding what AI agents are, their types, real-world use, and enterprise reality
- LLM, RAG, AI Agents, and Agentic AI – Key differences and how they connect
- Role, Goal, and Objectives – How agents think and act
- From Prompt Engineering to Context Engineering – Moving beyond prompt tweaks to structured reasoning
- Industry Use Cases – How different sectors are applying agentic systems
- Single-Agent Systems – Basics, setup, and limitations
- Multi-Agent Systems – Collaboration and task sharing between agents
- Core Architecture Patterns – ReAct, Reflection, Reflexion, Routing, and more
- Orchestration Patterns – Sequential, Hierarchical, Consensus, Map-Reduce, etc.
- Memory in Agentic Systems – How agents remember and use context effectively
- Code-Based Frameworks – LangGraph, CrewAI, AutoGen, Agents SDK, and hands-on examples
- No-Code Tools – n8n, OpenAI Agent Builder, and visual workflows
- Model Context Protocol (MCP) – A standard for context exchange between agents
- Evaluations and Metrics – How to measure performance and reliability
- Governance & Operations – Observability, monitoring, and safety practices
- The Future of Agentic AI – Trends, open standards, and what’s next
Key takeaways
Participants will learn how to design and build real AI agents that go beyond proof-of-concepts. They’ll gain practical insights from real production experiences—covering what actually works, what breaks, and how to deploy agent systems reliably.
What domain would you say your talk falls under?
Artificial Intelligence & Deep Learning
Duration (including Q&A)
60 minutes
Prerequisites and preparation
Basic knowledge of Python, a working laptop, stable internet connection, and curiosity to learn how real AI agents are built and deployed. No advanced AI background required.
Resources and references
Workshop Reference Repo
PyCon 2025 Repo
Link to slides/demos (if available)
https://www.youtube.com/watch?v=K5YwsoIUcgY
Twitter/X handle (optional)
No response
LinkedIn profile (optional)
https://www.linkedin.com/in/sonukr0
Profile picture URL (optional)
No response
Speaker bio
Sonu Kumar is a serial AI entrepreneur, YouTuber, and angel investor who has built and exited startups in the US. He’s passionate about advancing AI literacy and enabling a skilled AI-first ecosystem. Through his YouTube channel and the AI Anytime community, he’s helped thousands of learners and professionals understand, build, and grow with AI.
Availability
Third Saturday in November
Accessibility & special requirements
Accommodations/Logistics
Speaker checklist
- I have read and understood the PyDelhi guidelines for submitting proposals and giving talks
- I will make my talk accessible to all attendees and will proactively ask for any accommodations or special requirements I might need
- I agree to share slides, code snippets, and other materials used during the talk with the community
- I will follow PyDelhi's Code of Conduct and maintain a welcoming, inclusive environment throughout my participation
- I understand that PyDelhi meetups are community-centric events focused on learning, knowledge sharing, and networking, and I will respect this ethos by not using this platform for self-promotion or hiring pitches during my presentation, unless explicitly invited to do so by means of a sponsorship or similar arrangement
- If the talk is recorded by the PyDelhi team, I grant permission to release the video on PyDelhi's YouTube channel under the CC-BY-4.0 license, or a different license of my choosing if I am specifying it in my proposal or with the materials I share
Additional comments
No response