I’m an Engineer at Google. Here’s My Advice for New Engineers.

The tech industry is changing at an extraordinary pace, and it’s a truly exciting time to have a front-row seat to its biggest evolution yet.

Landing a job at Google is a dream for many engineers, including myself. I was thrilled to join the team two years ago as a software engineer. I came in with experience from other major tech companies, like Splunk and Amazon Web Services.

The world of Big Tech has become much more dynamic and results-oriented than the one I entered just a couple of years ago. The traditional software development cycle of long planning phases and multi-week sprints has been streamlined in favor of agility and faster execution. This shift has redefined priorities, required skills, and what it means to have an “impact.”

There are now more opportunities for engineers to contribute directly and meaningfully. Based on my experience, these are the most significant changes I’ve seen, and my advice for new engineers looking to thrive in this exciting environment.

AI isn’t just a feature anymore; it’s the foundation

The most profound change has been the explosion of generative AI. When I started my tech career in 2017, AI was a specialized field. Now, it’s a core competency that’s being integrated into everything. This has fundamentally changed the job of a software engineer.

The demand for expertise in machine learning systems and emerging areas like agentic AI systems has skyrocketed. You’re no longer just building an application; you’re expected to build the intelligent systems that power it.

In my own work, AI tools now shape how I debug, experiment, and optimize, turning workflows that once took days into hours. The expectation is clear: it’s not enough to build something that works — it has to be smart. The rules of data have fundamentally changed.

For an engineer, this changes the nature of your work. Building a “search” feature is no longer about simple keyword matching; it’s about creating a system that is smart and intelligent enough to know users’ needs. Additionally, ensuring that our intelligent systems are fair, ethical, and secure is critical in software development.

Build your core computer science knowledge

The most exciting new technologies are all built on timeless principles. Build your foundation on the bedrock of computer science: computer systems, distributed systems, software architecture, machine learning, artificial intelligence, and cloud systems. This core knowledge will outlast any specific tool.

Make it a habit to engage with academic research, follow conferences, and read papers. For example, I always keep an eye on the papers coming out of major AI conferences like NeurIPS, the Conference on Neural Information Processing Systems. I also follow the work of influential researchers like Andrej Karpathy.

This forward-looking curiosity is what helps me understand what is driving the next wave of products long before they become mainstream.

Go deep on AI: Be a builder and a power user

In today’s world, a surface-level understanding of AI isn’t enough. My advice is to go deep on AI from two angles. First, learn to be a builder. This means moving beyond simply calling an API and understanding the full stack of modern artificial intelligence. Dig into the foundations that power today’s most advanced models. The frontier is now multimodal, so a deep understanding of how models process and connect language, vision, and audio is essential.

Second, be a power user of AI tools. Proficiency in using AI for code generation, debugging, and rapid prototyping is no longer a luxury — it’s a massive force multiplier. This means moving beyond basic prompting and integrating the best available tools into every facet of your workflow.

The engineers who will define the next decade of software are those who master this duality: treating AI as both the raw material they build with and the intelligent tool that builds alongside them.

Turn knowledge into tangible projects

Theory is nothing without application. The most important step is to turn your knowledge into tangible projects. What truly matters is shipping a substantial product from start to finish.

For my personal project, to truly understand agentic systems, I built my own deep research agent from the ground up. It wasn’t a simple script but a multi-agent system built using open source frameworks like LangGraph. It featured a supervisor agent that would take a complex topic, decompose it into distinct sub-questions, and then delegate those questions to multiple researcher agents. Finally, the system would synthesize all the findings into a single report.

Building this project did more than just crystallize my learning; it created a powerful, tangible tool and served as a definitive example of my ability to architect and deliver complex AI systems. Don’t just learn; build and launch.

Harsh Varshney is a machine learning software engineer at Google. The views expressed here are his own and not Google’s.