Last updated:
Building a machine learning model that works on a laptop is one thing. Making it work reliably when data, users, and workloads grow is a very different challenge. This gap between experimentation and large-scale implementation was among the key subjects discussed by Mr. Siddhant Agarwal, Senior DevRel Advocate at ClickHouse, during Constellation 2026 at Polaris School of Technology.
His session brought students’ attention to a part of engineering that often gets less attention than model building itself: the infrastructure needed to make technology work at scale.
Mr. Agarwal also spoke about the changing role of AI in software development. For students preparing to enter the industry over the next few years, his message was not to treat AI as an easy way to finish tasks, but to understand it as an important part of the engineering process.
From a Small Machine Learning Model to a Scalable System
A machine learning project often begins in a controlled environment. A student writes code, trains a model, tests it with a dataset, and gets the expected output. The experience changes when the same system has to handle much larger amounts of data and real-world usage.
Questions around data storage, processing speed, system performance, and scalability become difficult to ignore.
Mr. Agarwal spoke about these challenges with students, explaining the difference between coding in an isolated setting and building infrastructure that can handle bigger workloads.
This is an important distinction for future engineers. A working prototype is a good start, but it doesn’t automatically make a system ready for real-world use. Understanding what happens as the system grows is part of the job.
Why Data Infrastructure Deserves More Attention?
Machine learning and AI discussions often revolve around models, algorithms, and their outputs. But behind every useful AI application is an infrastructure layer that manages data and makes information available when it is needed.
The volume of data is increasing, and systems need to be able to process it efficiently without degrading performance. Engineers must think about how data is handled, where bottlenecks will occur, and what changes are required as workloads increase.
This is not just a problem for the big tech companies. Even students on personal projects can learn to think about them early.
Mr. Agarwal’s talk challenged students to think beyond whether their code works, and instead to consider if the system they have built can handle growth.
AI Should Be More Than a Shortcut
Another important aspect of Agarwal’s address was the role of AI in the work of an engineer.
AI tools can help developers write code, understand errors, and complete some tasks faster. But using them only to save time misses the larger opportunity.
The way engineers work with AI is changing. It is becoming part of the technical process rather than remaining a tool used only when someone needs a quick answer. Students need to understand how these systems work, where they can be useful, and how to evaluate the results they produce.
There is also a difference between accepting a solution generated by AI and knowing that it is correct. Engineers still have to review code, think about security and performance, and be responsible for the final output.
As development practices keep evolving, we will have to learn to work with AI, instead of blindly depending on it.
What Should Engineering Students Learn Now?
Students entering college today will graduate into a technology industry that may look quite different from the one they see now. Tools will change, new frameworks will emerge, and AI will continue to influence software development.
Agarwal’s message was a reminder that once you’ve learned a programming language or finished a few projects, students can’t afford to stop learning.
They need to know data systems, scalability, and the technical underpinnings of AI applications inside out. It’s also important to know how to question the output of AI tools, to not take every suggestion as a ready-made solution.
For students, one way to implement this advice is to take an existing project and ask a few questions:
- If the data volume increases, what happens?
- Will the application still work well with more users?
- Where could the system slow down?
- Can AI tools help improve the development process, and how will their output be checked?
- What changes would be needed before the project could be used in a real-world setting?
These questions can turn a basic project into a much better learning experience.
Key Takeaways From Siddhant Agarwal’s Address at Constellation 2026
Siddhant Agarwal’s session emphasised that engineering is more than simply writing code or building a model that works in a test environment. Students need to understand the infrastructure that underlies technology, how systems work at scale, and to be able to engage with AI thoughtfully. His message was particularly relevant for those getting ready to join an industry in which data and AI are increasingly at the center of software development.
Building Engineers for the Next Phase of Technology
Agarwal’s talk at Constellation 2026 offered students a view of the challenges that follow the initial excitement of building a new technology. Getting a model to work is an accomplishment, but to make it reliable, efficient, and useful at scale is another order of engineering.
No difference with AI. Knowing how to use an AI tool may help students complete a task today, but knowing how to work with AI as part of a larger technical system will be far more valuable in the long run.
Students at Polaris School of Technology take away that they should remain curious, experiment on real projects, and pay attention to what happens behind the code. The engineers who understand both the technology and the systems that support it will be better prepared for the demands of the coming years.