Because my formal education was not in computer science or artificial intelligence, most of my understanding of the field has come through self-teaching. I have learned by following questions beyond the point at which the easy explanations end—reading, building, getting things wrong, and returning to the fundamentals with a clearer sense of what I did not understand.
Over time, I have worked across computer vision, RAG systems, agentic AI, personalization, context engineering, cloud deployment, and local intelligence. At IIT Bombay’s AI community, I was part of a small team building an AI agent that used retrieval-augmented generation to make institute-specific information accessible to students. I later worked on image-similarity search and optical character recognition at Big Vision, explored agentic AI and personalization at Motilal Oswal Financial Services, contributed to the core AI systems at SuperLiving during its early stage, and worked on the idea of local intelligence through Srota before joining Meesho.
From models to complete systems
These experiences have gradually shifted my interest from individual models toward the architecture of complete AI systems. I enjoy understanding how different systems are put together: how data and context move through them, how their components communicate, where latency accumulates, how they behave under failure, and which architectural choices determine whether a promising prototype can become a reliable product.
The next stage of my technical exploration moves in two directions. I want to go one level above my current abstraction and learn how to build systems that scale: distributed systems, infrastructure, reliability, observability, data movement, and the decisions required to serve real users. At the same time, I want to go one level below—to understand CPUs and GPUs, memory hierarchies, inference engines, kernels, compilers, and the optimizations that determine how efficiently a model runs.
Eventually, I want to follow an AI system across the entire stack—and upward again into a dependable product used by millions of people.
Research directions
My current research interests include test-time training, continual learning, and RDL—particularly the search for something resembling a “bitter truth” for data science. I am interested in systems that do not remain frozen after training but can adapt to new information, changing environments, and the people using them.
Edge deployment sits at the intersection of much of this curiosity. I believe that within five years, models more capable than many of today’s state-of-the-art systems will be able to run locally on ordinary phones. That possibility interests me because intelligence could become more private, personal, accessible, and less dependent on centralized infrastructure.
Entrepreneurship and the wider world
The next gap I want to close is not technical. I want to understand how businesses in the technology industry actually work: how a technical insight becomes a product, how companies recognize genuine demand, why distribution can matter as much as invention, and how teams balance speed, quality, capital, and conviction.
I hope to start a company in the future. To me, entrepreneurship is the difficult process of taking an idea out of your head, exposing it to reality, and making it useful enough that other people willingly bring it into their lives. I want to build something technically meaningful, commercially sustainable, and honest about the problem it is trying to solve.
Outside technology, I enjoy chess, UFC, geopolitics, and history. Each reveals something about decision-making under pressure—how people adapt to constraints, act with incomplete information, respond to incentives, and behave when the consequences are real. My interests in neuroscience and psychology come from the same underlying curiosity.
Personally, the next chapter of my life is also about becoming more physically grounded. I want to take fitness seriously, travel widely, and encounter ways of living and thinking that cannot be understood from a screen.
Why this website exists
I do not want this website to present me as a finished person or turn my life into an expanded résumé. I want it to be a living record of how I think: what I am learning, what I currently believe, where I remain uncertain, and how my ideas change when they encounter better evidence.
Some thoughts may become essays, while others may remain learning notes or unfinished research ideas. The development of the thinking matters to me as much as the final conclusion.