Building Data Solutions People Want to Use

My path into data engineering didn’t start with data.

I studied industrial engineering at Afeka Tel Aviv Academic College of Engineering, specializing in information systems, and my first professional role was as a BI Developer at Elbit Systems. I worked closely with product and engineering teams and communicated directly with end users who relied on my work’s outputs to analyze data and gain reliable insights.

I remember how valuable their feedback was, and how rewarding it was to see their enthusiasm when a new feature truly met their needs. It was a role that taught me something I still carry today: good technology creates value only when it works for the people using it.

As part of this role, I built both data infrastructure and analytical solutions. Over time, I found myself wanting to go deeper into the technical side of how those systems worked. That led me to data engineering.

The goal, however, stayed the same. I wanted to build data solutions that maximize value for end users while minimizing the effort required to use them. The best products, in my view, are reliable, scalable, and intuitive enough that people actually want to use them.

Growing Alongside a Changing Field

I’ve been at Tipalti for three and a half years. I started as a data engineer and later became a senior data engineer, and my role has evolved considerably along the way.

Today, as part of the AI and Data Platform team, my work spans two areas. One is building the data pipelines and AI infrastructure that enable teams across the organization to develop and deploy AI-powered data products efficiently. The other is building AI-driven data solutions directly for end users.

That evolution reflects how quickly the field itself is changing. Agentic AI is becoming part of everyday work, not just something we experiment with on the side. It’s changing how we operate, how we build, and how people interact with data.

For me, that’s where data engineering becomes especially interesting. It’s no longer only about moving data from one system to another. It’s about creating the infrastructure, context, and tools that allow people to ask better questions and get useful answers faster.

Turning Conversations Into Business Insights

One of the technology initiatives I’ve led at Tipalti was the development of an Account Insights and Interactions Agent designed to make the organization’s data more accessible and useful.

• The Challenge: The idea was straightforward, but technically complex: give authorized internal employees one place to analyze structured data, such as enriched business data, alongside unstructured data, such as conversation transcripts. We wanted to create a unified experience where those different sources could work together, rather than requiring people to manually piece together information from multiple places.
  
• The Solution: We combined two components. The first is a semantic model that encodes business definitions, including metrics, relationships, and terminology. This allows a text-to-SQL layer to translate natural-language questions into accurate, governed queries over our enriched business data. The second is a retrieval (RAG) layer over interaction transcripts. We built a dedicated pipeline to create a 360-style interactions table, bringing all interactions together in one place. We also designed it to be AI-ready, so the data is ready for chunking and embedding, enabling semantic search based on meaning rather than just keywords.
  
  An orchestration layer determines which source is needed for each question, or whether both are required, and combines the results into a single answer. Role-based access controls ensure that each employee can only access the data they are authorized to see.
  
• The Results: More than 290 authorized internal employees have used the Agent, with an 83% repeat rate and an average of more than 15 sessions per user. Those numbers tell me something important: technology is most valuable when it fits naturally into how people already work. The Agent isn’t replacing the human analysis or business knowledge behind those questions. It’s making that knowledge easier to apply, giving people a faster way to explore information and turn it into useful business insights.

Building Faster and Smarter with AI

My own experimentation with AI has made one thing very clear: it can significantly increase development velocity. It allows organizations to build self-service solutions faster and bring useful capabilities to more people.

But speed also creates a new challenge. I think of it as “insights inflation.” If everyone can ask AI to generate an analysis, the same question can be answered repeatedly by different people. That can lead to duplicated effort, inconsistent interpretations, and unnecessary compute costs.

That doesn’t mean we should slow down innovation. Quite the opposite. It means we need to give innovation the right foundation.

I believe organizations should work toward a governed, centralized analytics hub that handles the majority of common analytical needs. It can provide trusted, reusable capabilities such as predefined insights, skills, tools, and AI agents, while still giving people the flexibility to create their own solutions for more specialized questions.

That balance is important. We can use AI to make data more accessible without losing the governance and context that make the data trustworthy in the first place.

The Human Side of Data

If I’ve learned anything throughout my career, it’s that technical experience is only part of the job. You have to stay curious.

My background has taken me from BI development to data engineering and now into AI. Each transition required me to learn something new and be willing to rethink how I approached problems to make people’s jobs easier. That mindset has been just as important as the technology itself.

AI makes that human element even more important. The technology can help us move faster, explore more possibilities, and build things that would have taken much longer before. But we still need people who can ask the right questions, understand business context, communicate clearly, and recognize when an answer is actually useful.

For other data engineers, my advice is simple: don’t be afraid of change. Embrace it.

The technologies we work with will keep evolving. AI is already changing what it means to be a data engineer, just as data engineering changed what came before it. Those changes can feel disruptive, but they are also opportunities to learn, take ownership, and stand out.