AI REWIND 2025 - MLOps Reading Group Year-end Special | MLOps.community

The podcast features a panel of AI experts dissecting consequential AI developments from the past year, categorized into architecting AI workflows, multi-agent orchestration, and model cognition. Discussions cover the evolution of AI agents from experimental demos to production systems, emphasizing the importance of control, tools, memory, and guardrails. The rise of white coding is explored, noting its benefits for rapid prototyping but cautioning against its use in production due to potential risks. Context engineering is presented as a shift towards precision in providing information to models, contrasting it with traditional prompt engineering. The panel also discusses the Model Context Protocol (MCP) and continuous evaluation frameworks, highlighting the need for security and adaptability in AI systems.

Outlines

Part 1: Introduction and AI Agent Fundamentals

Part 2: Context Engineering and Standardization

Part 3: Evaluation and Orchestration Strategies

Part 4: Model Training and Alignment

Part 5: Real-World Validation and Market Trends

Part 6: Architecture and Future Outlook

Sign in to continue reading, translating and more.

Open full episode in Podwise

AI REWIND 2025 - MLOps Reading Group Year-end Special

MLOps.community

Part 1: Introduction and AI Agent Fundamentals

Introduction to AI Consequential Developments and Speaker Introductions

AI Agents: Transitioning from Experimentation to Production with Guardrails

The Rise of White Coding: Steering AI for Rapid Prototyping

Part 2: Context Engineering and Standardization

Context Engineering: Prioritizing Precision over Recall for Effective LLM Performance

MCP: Standardizing AI Model Access to Tools and Data Sources

MCP Adoption, Security Risks, and Implementation

Part 3: Evaluation and Orchestration Strategies

Continuous Evaluation Frameworks: Superseding One-Time Benchmark Validation

Context Engineering and MCP Discussion

Multi-Agent Orchestration: Research and Production Implementations

Part 4: Model Training and Alignment

Building ChatGPT: Reinforcement Learning and Human Feedback

Post-Training Techniques: SFT, DPO, KTO, PPO, and GRPO

Part 5: Real-World Validation and Market Trends

Product-Level Validation: Bridging the Gap Between Benchmarks and Real-World Success

The Openness of AI Models: How Good and How Adopted?

Part 6: Architecture and Future Outlook

Architectural Changes and Training Improvements in AI Models

Q&A: Domain-Specific Evaluation, Multi-Agent Systems, and GRPO Implementation

AI REWIND 2025 - MLOps Reading Group Year-end Special

MLOps.community

Part 1: Introduction and AI Agent Fundamentals

00:00Introduction to AI Consequential Developments and Speaker Introductions

Introduction to AI Consequential Developments and Speaker Introductions

02:01AI Agents: Transitioning from Experimentation to Production with Guardrails

AI Agents: Transitioning from Experimentation to Production with Guardrails

09:22The Rise of White Coding: Steering AI for Rapid Prototyping

The Rise of White Coding: Steering AI for Rapid Prototyping

Part 2: Context Engineering and Standardization

14:56Context Engineering: Prioritizing Precision over Recall for Effective LLM Performance

Context Engineering: Prioritizing Precision over Recall for Effective LLM Performance

26:25MCP: Standardizing AI Model Access to Tools and Data Sources

MCP: Standardizing AI Model Access to Tools and Data Sources

32:31MCP Adoption, Security Risks, and Implementation

MCP Adoption, Security Risks, and Implementation

Part 3: Evaluation and Orchestration Strategies

40:26Continuous Evaluation Frameworks: Superseding One-Time Benchmark Validation

Continuous Evaluation Frameworks: Superseding One-Time Benchmark Validation

48:24Context Engineering and MCP Discussion

Context Engineering and MCP Discussion

52:57Multi-Agent Orchestration: Research and Production Implementations

Multi-Agent Orchestration: Research and Production Implementations

Part 4: Model Training and Alignment

1:04:32Building ChatGPT: Reinforcement Learning and Human Feedback

Building ChatGPT: Reinforcement Learning and Human Feedback

1:11:08Post-Training Techniques: SFT, DPO, KTO, PPO, and GRPO

Post-Training Techniques: SFT, DPO, KTO, PPO, and GRPO

Part 5: Real-World Validation and Market Trends

1:20:41Product-Level Validation: Bridging the Gap Between Benchmarks and Real-World Success

Product-Level Validation: Bridging the Gap Between Benchmarks and Real-World Success

1:36:08The Openness of AI Models: How Good and How Adopted?

The Openness of AI Models: How Good and How Adopted?

Part 6: Architecture and Future Outlook

1:42:25Architectural Changes and Training Improvements in AI Models

Architectural Changes and Training Improvements in AI Models

1:53:30Q&A: Domain-Specific Evaluation, Multi-Agent Systems, and GRPO Implementation

Q&A: Domain-Specific Evaluation, Multi-Agent Systems, and GRPO Implementation