Intern interview & blog: Lillian Liu’s embrace of high-quality data infrastructure

Editor’s note: Lillian Liu was a summer 2025 intern supporting the energy, chemicals and infrastructure industries at Siemens Digital Industries Software.

Watch this interview with Lillian and read her blog post below describing her internship experience and skills he obtained.

---

Seeing the bigger picture: My internship journey with data (By Lillian Liu)

Data is everywhere. It’s the backbone of modern innovation, the fuel for AI, and the foundation of digital transformation. But as I’ve learned during my internship at Siemens, having data isn’t the same as having good data or using it well.

I’ve always known that good data matters. I remember sitting in AP Statistics, learning how sampling bias could skew results, and how using outdated or irrelevant data could completely distort an analysis. But it wasn’t until this internship that I truly understood the depth of data’s importance—and the cost of getting it wrong.

The reality check

In one of my first meetings, John Nixon, who led and mentored our team during the summer, said something that stuck with me:

“I’ve sat in shops with engineers where 90% of their time is spent looking for data instead of being creative with their designs. The other 10%? Going on breaks and complaining they can’t find the right document.”
—John Nixon, VP, Global Strategy, Energy Chemicals and Infrastructure

He admitted he was exaggerating a bit—but only slightly. He went on to reflect on how far we’ve come since the early days of digital file sharing, when just being able to store and retrieve files over a network was revolutionary. And yet, despite all that progress, we still spend an incredible amount of time just trying to find the data we need.

That perspective surprised me. Our generation is surrounded by data, how could it still be so hard to access the right information?

Then I started my first project.

I was tasked with modifying a Rolling Stock train model in NX software for use in technical marketing. Sounds straightforward, right? But I spent nearly half of my first day just trying to figure out what I was even looking for. Was it a coupler? A connector? A wheelset? Eventually, I narrowed it down to a Jacobs bogie and CASNUB bogie. And even then, it took another few days before we discovered the exact name of the cart: a Well car.

That experience was eye-opening. We’re drowning in data, but without structure and clarity, it’s like trying to find a needle in a haystack.

The cost of bad data

As I dug deeper, I learned just how expensive poor data quality can be. According to a 2021 Gartner report, bad data costs organizations an average of $12.9 million per year. In supply chain operations, this can mean excess inventory, delivery delays, stockouts, or added fuel costs.

That’s why I was excited to work on a project that directly addressed this issue: building an Inspection and Test App. The goal was to standardize inspection and testing processes across facilities, allowing data to be:

• Collected in a single, organized interface
• Analyzed immediately using Altair’s High Performance Computing (HPC)
• Stored long-term as a digital twin for simulation and optimization

Connecting the dots: Altair, Siemens, and the power of our integration

As I worked on the app, I became more curious about Altair and its role in the Siemens’ ecosystem. I even participated in a research project on integrating Altair’s software in the energy, chemicals and infrastructure projects.

That’s when it all clicked.

Simulations? They’re just math equations modeling real-world systems. But to simulate something accurately, you need data about that system: dimensions, materials, temperatures, flow rates. No data, no simulation.

AI? Same story. Early AI was logic-based, but it hit roadblocks like the frame problem^¹ and the common sense knowledge bottleneck.^² Today’s AI, especially large language models, succeeds because it’s trained on massive datasets. The quality of the data determines the quality of the model.

That’s why Siemens’ acquisition of Altair was so strategic. Siemens’ Xcelerator platform has vast amounts of industrial data. Altair brings the analytics and HPC power to process it. And Mendix, Siemens’ low-code platform, ties it all together—letting people build apps (like mine!) that collect, store, analyze, and present data seamlessly.

Together, this ecosystem doesn’t just make data accessible—it makes it actionable. For data-driven companies like Siemens, and many of our customers and potential partners, this integration empowers teams to make faster, more informed decisions. With analytics returning results more quickly and accurately, the true value of data emerges—in the answers it enables.

Looking back, moving forward

This internship was more than just a learning experience: it’s been a mindset shift. I’ve come to see data not just as numbers or files, but as the infrastructure of innovation. And I’ve seen firsthand how Siemens empowers interns like me to explore, contribute, and grow.

I’m grateful for the freedom I’ve had to choose meaningful projects, the mentorship I’ve received, and the chance to work on something that truly matters. I hope to continue this journey—and keep building a world where data works for us, not against us.

Footnotes

¹ Frame Problem: A challenge in early AI where systems struggled to understand what didn’t change in a situation. For example, if a cup is moved, the AI might not know that its color and shape stayed the same—unless every detail is explicitly coded. This made reasoning overly complex and inefficient.

² Common Sense Knowledge Bottleneck: Many things humans consider “common sense” don’t follow strict logic, making it hard for computers to grasp. Teaching AI these intuitive truths is difficult because they’re often implicit and context-dependent.