I am Tanvi, a Mechanical Engineering student at IIT Kanpur. I spend most of my time thinking about intelligence and systems rather than machines. What fascinates me are the hidden layers of software and AI, the parts that are easy to overlook but that quietly shape how everything works.
My main interest is AI systems design. Current AI can generate fluent answers, but it cannot reliably decide what to believe or carry knowledge forward once a session ends. I keep coming back to two missing ideas: credibility and continuity. Credibility is about teaching AI to reason with evidence rather than treating all information as equal. Continuity is about giving AI memory so that every interaction does not begin from zero.
I explore these questions through epistemic frameworks, layered memory, and multi-agent systems. I like to study how autonomous components might coordinate, share tasks, and adapt in situations where centralized systems fail. Much of what I build are prototypes and small experiments that test assumptions. They are rarely polished, but they help me see how intelligence might scale if built on stronger foundations.
I am also a founder-in-the-making. In 2025, I was selected into Entrepreneurs First (EF). I chose to defer my participation for a year so that I can finish my degree and focus fully afterward. Right now I am working in stealth on Coddbase, a deep-tech startup that is trying to create the foundations of trustworthy intelligence.
This site is my space to write about the questions that drive me. Sometimes I post long essays that connect ideas across systems. Other times I capture quick notes so that half-formed thoughts do not slip away. My posting is uneven, but it reflects the rabbit holes I follow when something catches my attention.
If you are curious about how AI might learn to remember, to reason with evidence, and to evolve with time, or if you enjoy rethinking the defaults hidden in the tools we use every day, you may find something here that resonates.