POC vs Production: Compounding Errors and Reliability Challenges

Summary
In this episode of the Data Engineering Podcast Omri Lifshitz (CTO) and Ido Bronstein (CEO) of Upriver talk about the growing gap between AI's demand for high-quality data and organizations' current data practices. They discuss why AI accelerates both the supply and demand sides of data, highlighting that the bottleneck lies in the "middle layer" of curation, semantics, and serving. Omri and Ido outline a three-part framework for making data usable by LLMs and agents: collect, curate, serve, and share challenges of scaling from POCs to production, including compounding error rates and reliability concerns. They also explore organizational shifts, patterns for managing context windows, pragmatic views on schema choices, and Upriver's approach to building autonomous data workflows using determinism and LLMs at the right boundaries. The conversation concludes with a look ahead to AI-first data platforms where engineers supervise business semantics while automation stitches technical details end-to-end.


Announcements

• Hello and welcome to the Data Engineering Podcast, the show about modern data management
• Data teams everywhere face the same problem: they're forcing ML models, streaming data, and real-time processing through orchestration tools built for simple ETL. The result? Inflexible infrastructure that can't adapt to different workloads. That's why Cash App and Cisco rely on Prefect. Cash App's fraud detection team got what they needed - flexible compute options, isolated environments for custom packages, and seamless data exchange between workflows. Each model runs on the right infrastructure, whether that's high-memory machines or distributed compute. Orchestration is the foundation that determines whether your data team ships or struggles. ETL, ML model training, AI Engineering, Streaming - Prefect runs it all from ingestion to activation in one platform. Whoop and 1Password also trust Prefect for their data operations. If these industry leaders use Prefect for critical workflows, see what it can do for you at dataengineeringpodcast.com/prefect.
• Data migrations are brutal. They drag on for months—sometimes years—burning through resources and crushing team morale. Datafold's AI-powered Migration Agent changes all that. Their unique combination of AI code translation and automated data validation has helped companies complete migrations up to 10 times faster than manual approaches. And they're so confident in their solution, they'll actually guarantee your timeline in writing. Ready to turn your year-long migration into weeks? Visit dataengineeringpodcast.com/datafold today for the details.
• Composable data infrastructure is great, until you spend all of your time gluing it together. Bruin is an open source framework, driven from the command line, that makes integration a breeze. Write Python and SQL to handle the business logic, and let Bruin handle the heavy lifting of data movement, lineage tracking, data quality monitoring, and governance enforcement. Bruin allows you to build end-to-end data workflows using AI, has connectors for hundreds of platforms, and helps data teams deliver faster. Teams that use Bruin need less engineering effort to process data and benefit from a fully integrated data platform. Go to dataengineeringpodcast.com/bruin today to get started. And for dbt Cloud customers, they'll give you $1,000 credit to migrate to Bruin Cloud.
• Your host is Tobias Macey and today I'm interviewing Omri Lifshitz and Ido Bronstein about the challenges of keeping up with the demand for data when supporting AI systems

Interview

• Introduction
• How did you get involved in the area of data management?
• We're here to talk about "The Growing Gap Between Data & AI". From your perspective, what is this gap, and why do you think it's widening so rapidly right now?
• How does this gap relate to the founding story of Upriver? What problems were you and your co-founders experiencing that led you to build this?
• The core premise of new AI tools, from RAG pipelines to LLM agents, is that they are only as good as the data they're given. How does this "garbage in, garbage out" problem change when the "in" is not a static file but a complex, high-velocity, and constantly changing data pipeline?
• Upriver is described as an "intelligent agent system" and an "autonomous data engineer." This is a fascinating "AI to solve for AI" approach. Can you describe this agent-based architecture and how it specifically works to bridge that data-AI gap?
• Your website mentions a "Data Context Layer" that turns "tribal knowledge" into a "machine-usable mode." This sounds critical for AI. How do you capture that context, and how does it make data "AI-ready" in a way that a traditional data catalog or quality tool doesn't?
• What are the most innovative or unexpected ways you've seen companies trying to make their data "AI-ready"? And where are the biggest points of failure you observe?
• What has been the most challenging or unexpected lesson you've learned while building an AI system (Upriver) that is designed to fix the data foundation for other AI systems?
• When is an autonomous, agent-based approach not the right solution for a team's data quality problems? What organizational or technical maturity is required to even start closing this data-AI gap?
• What do you have planned for the future of Upriver? And looking more broadly, how do you see this gap between data and AI evolving over the next few years?

Contact Info

• Ido - LinkedIn
• Omri - LinkedIn

Parting Question

• From your perspective, what is the biggest gap in the tooling or technology for data management today?

Closing Announcements

• Thank you for listening! Don't forget to check out our other shows. Podcast.__init__ covers the Python language, its community, and the innovative ways it is being used. The AI Engineering Podcast is your guide to the fast-moving world of building AI systems.
• Visit the site to subscribe to the show, sign up for the mailing list, and read the show notes.
• If you've learned something or tried out a project from the show then tell us about it! Email hosts@dataengineeringpodcast.com with your story.

Links

• Upriver
• RAG == Retrieval Augmented Generation
  • AI Engineering Podcast Episode
• AI Agent
• Context Window
• Model Finetuning)

The intro and outro music is from The Hug by The Freak Fandango Orchestra / CC BY-SA