Generative AI is now a core competency in software engineering, transforming the field.<\/p>\n
AI changed the way we use data and turned workflows that once took days into hours.<\/p>\n
Engineers must master AI as both a building material and a tool for future software.<\/p>\n
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.<\/p>\n
Landing a job at Google<\/a> 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.<\/p>\n 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.”<\/p>\n 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<\/a> looking to thrive in this exciting environment.<\/p>\n AI isn’t just a feature anymore; it’s the foundation<\/p>\n 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.<\/p>\n The demand for expertise in machine learning<\/a> 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.<\/p>\n 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 \u2014 it has to be smart. The rules of data have fundamentally changed.<\/p>\n For an engineer, this changes the nature of your work. Building a “search” feature is no longer about simple keyword matching<\/a>; 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.<\/p>\n Build your core computer science knowledge<\/p>\n 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<\/a>, artificial intelligence, and cloud systems. This core knowledge will outlast any specific tool.<\/p>\n 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.<\/p>\n This forward-looking curiosity is what helps me understand what is driving the next wave of products long before they become mainstream.<\/p>\n Go deep on AI: Be a builder and a power user<\/p>\n 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<\/a> 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.<\/p>\n Second, be a power user of AI tools. Proficiency in using AI for code generation, debugging, and rapid prototyping is no longer a luxury \u2014 it’s a massive force multiplier. This means moving beyond basic prompting and integrating the best available tools into every facet of your workflow.<\/p>\n The engineers who will define the next decade of software<\/a> are those who master this duality: treating AI as both the raw material they build with and the intelligent tool that builds alongside them.<\/p>\n Turn knowledge into tangible projects<\/p>\n 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.<\/p>\n 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<\/a> 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.<\/p>\n 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<\/a>. Don’t just learn; build and launch.<\/p>\n Harsh Varshney is a machine learning software engineer at Google. The views expressed here are his own and not Google’s.<\/p>\n