I'm Tanvi, a mechanical engineering student at Indian Institute of Technology Kanpur (IIT-K), who spends a lot of time thinking about the invisible infrastructure that shapes how we build software. My curiosity lies in the messy, overlooked systems that run behind the scenes and how reimagining them could fundamentally change our daily work.
My research interests center on Distributed Multi-Agent Systems and their applications. I'm designing scalable, intelligent agent architectures where autonomous components collaborate and delegate tasks to solve complex problems more efficiently than traditional centralized systems. This work spans multiple domains: improving developer tools, advancing AI infrastructure, enabling dynamic robotics, and streamlining software workflows by addressing core bottlenecks like context limits, coordination overhead, and real-time adaptability.
A key focus of my research involves integrating layered memory and stateful reasoning into agent frameworks. This enables long-term planning, persistent context retention, and intelligent conflict resolution in distributed environments, whether that's coordinating robotic limbs that make split-second decisions or managing AI agents that maintain coherent state across complex software projects.
Currently, I'm building several interconnected projects that explore different facets of this vision. I'm reimagining continuous integration to make builds smarter and faster through AI-driven optimization and performance SLAs. I'm developing systems that automatically generate GitHub workflows by understanding project modularity and dependencies, essentially creating a workflow architect that optimizes for incremental builds and parallel execution.
I'm also working on a universal AI memory infrastructure that allows tools to remember context across sessions and devices, enabling truly stateful AI interactions. Alongside this, I'm exploring privacy-first communication platforms with real-time AI moderation that puts control back in users' hands. Then there's my AI-powered coding environment that uses multi-agent systems and hierarchical memory to overcome the context limitations and coordination challenges in current development tools.
None of these are polished products yet. They're experiments, early prototypes, and sometimes questions encoded in code. Each project pushes against assumptions baked into the tools we use daily, questioning the defaults, configurations, and protocols that often go unexamined. They're all connected by a common thread: the belief that distributed intelligence can solve problems that centralized systems struggle with.
This website is where I step back from the code to explore the "why" behind these systems. It's my space to document the curiosities, frustrations, and breakthroughs that emerge when you look closer at how things actually work versus how we assume they work. Sometimes that means writing detailed technical essays, sometimes it's quick notes or sketches of ideas still taking shape.
Fair warning: I'm not always consistent with posting. Sometimes I go quiet for weeks, then suddenly flood the blog with multiple posts when a bunch of half-formed thoughts catch up and demand to be written down before they slip away. If you see a burst of activity, that's me chasing down the rabbit holes I left open.
The intersection of mechanical systems thinking and software infrastructure has led me to some unexpected insights about how intelligence scales, how systems coordinate, and how we might build tools that actually understand the work we're trying to do.
A lot of cool developments are coming up as these projects mature and the connections between them become clearer.
If you're curious about the hidden layers beneath your tools, or if you find yourself wondering why something works the way it does especially when it feels a bit off you'll probably find something here that resonates.