
ALT My afternoons are all about collaboration! After lunch, I met with an enterprise customer’s technical team to walk through architecture designs and answer their questions. Whenever complex edge cases arise, I bring in specialists across data and infrastructure to whiteboard solutions and pressure-test our designs.

ALT To finish the day, I jumped into another dedicated work block for hands-on problem-solving. In my role, I don't just advise — I build. I used this time to spin up proof-of-concepts in the Google Cloud Console and test workload scalability. Before wrapping up, I reviewed final deliverables, sent updates to key stakeholders, and set my priorities for tomorrow.

ALT "My project provides a generic framework that autonomously tunes operating system parameters for different AI workloads while optimizing critical compute resources like memory. Traditionally, kernel engineers pore over system telemetry, make initial guesses, and rely on standard statistical machine learning models to optimize parameters. This is a slow process that can lead to spiky, unstable system performance. By offloading this tuning to autonomous agents, we can implement better policies to support these workloads."

ALT Hi! I’m Sumit, a Cloud Architect with Google Cloud’s Global Solutions Delivery (GSD) team. My role focuses on helping Google's enterprise customers design and build scalable, secure cloud and AI architecture. Here’s what a typical day looks like!

ALT My morning starts with a gCab ride into the office. While commuting, I catch up on emails and chats so I can prioritize my schedule. Once I arrive at the office, I head straight to grab coffee with my teammates, where we catch up before diving into deep work. The core of my role is about problem-solving, so I spend a large chunk of my day designing, reviewing, and implementing scalable cloud architectures for Google Cloud customers.

ALT Around mid-day, I grab lunch and a quick recharge with a friendly game of table tennis or pool with fellow Googlers. Afterwards, my afternoon is packed with collaboration. I start with an internal standup with our Google team to align on customer project deliverables, followed by an external sync with the customer’s technical team to review architecture designs, address questions, and ensure seamless execution.

ALT "Buildings consume a massive share of global energy and emissions, making them a prime target for efficiency. However, because every building is unique, scaling reinforcement learning has been difficult... To solve this, we use a custom open-source simulator to build an accurate digital twin of the building. We then use behavior cloning to mimic current conditions, refine it inside the digital twin to optimize both comfort and energy savings, and deploy it with built-in guardrails to ensure compliance with state regulations."




























