Agenda & Outcomes¶
Agenda¶
- Introduction to Intelligent App Development with Copilot stack
- Unveiling the Art of the Possible: Inspiring Walkthroughs
- Comprehensive Architecture Overview
- Essential Pre-requisites
- Unearthing Customer Needs and Pain Points (Design Thinking session to personalize the workshop for your organization needs and use cases)
- Crafting AI-Driven Design and User Experience
- First steps to integrate AI into cloud-native and Azure backing services and data processing
- Empowering Development with AI Capabilities (inner-loop), including leveraging Github CoPilot
- Learn LLM capabilities and Prompt Engineering
- In-context learning (ICL)
- Prompt Engineering techniques
- Advanced Retrieval augmented generation (RaG)
- Integrating vector databases
- Prompt Flow
- Agents with Semantic Kernel, Autogen, and TaskWeaver
- Deep-dive into AI orchestration primitives through use cases with Miyagi and Reddog
- Exploring Agents and Agent-like Orchestration
- Tradeoffs with AI orchestrators (Semantic Kernel, Langchain, Guidance, TypeChat, PromptFlow etc.)
- Copilot Studio and Power Platform
- AI Studio and Prompt Flow
- AI Content Safety Studio
- Fine-tuning OSS models with proprietary training data to improve task specific accuracy and latency
- Streamlined Deployment and Monitoring of AI-Infused Apps
- (Optional) First Principles of Large Language Models for application developers
- High Level Overview of AI and ML
- Overview and History of NLP, NLU, and NLG
- Autoregressive Models
- NLP: RNNs to Transformers
- Computer Vision: CNNs to Vision Transformers (ViT)
- Audio: Wav2Vec2, Whisper
- Multimodal: VisualBERT, CLIP
- Reliability & Responsible AI for developers
- Ensuring Reliability and Controllability of LLM Output
- Grounding
- Alignment
- Retrieval augmented generation (RaG)
- Ensuring Reliability and Controllability of LLM Output
- Hands-on Project: Constructing a Semantic Kernel Plugin and integrating it with an Existing App using Miyagi
- Reliability and controllability of LLM output
- Embracing "Everything as Code": Automation, Deployment, and Operationalization for production
- Architecture Design Session: Crafting a customized PoC for your unique use case(s)
- Conclusion and Next Steps with MVP
Outcomes¶
Upon completing this workshop, participants will:
- Comprehend the transformative potential of Azure AI services and Copilot Stack in revolutionizing every facet of AI app development.
- Acquire hands-on experience in AI-driven design and user experience techniques
- Learn to effortlessly integrate foundation models into backend services, data processing, and frontend development
- Build an AI-infused app from the ground up during the hands-on project
- Implement Agents using Assistants API, Code Interpretor, and Retrieval intefaces.