MLOps.community AI Reliability, Spark, Observability, SLAs and Starting an AI Infra Company
Jun 27, 2025
Kostas Pardalis and Yoni Michael, co-founders of Typedef, dive into the transformative power of AI and data in today’s business landscape. They discuss how LLMs are revolutionizing data pipelines and the vital role of AI Reliability Engineers in ensuring system stability. The duo emphasizes the need for robust infrastructure and collaboration between engineers and domain experts to tackle integration challenges. They also explore the evolving dynamics of AI in customer support and the importance of community in driving innovation in technology.
AI Snips
Chapters
Transcript
Episode notes
Legacy Data Tools Limit AI
- Existing data tools were designed over a decade ago for structured BI workloads and aren't fit for modern AI and unstructured data use cases.
- New AI workloads require tooling built with first principles for unstructured and multimodal data processing.
AI Catalyzes Data-as-Product
- AI accelerates the evolution of data into a product, against a backdrop of mature SaaS software platforms.
- Unstructured data and AI models now form the core data products, requiring a rethink of data infrastructure.
Gemini LLM Failure Story
- Demetrios shared how his Gemini LLM task failed midway, losing progress and costing resources.
- This highlighted the unreliability challenge when running large contextual LLM requests in production.