Hi, I'm Shubhayan, studying CS at USC, where I spend most of my time in product, engineering and design. Currently, I work at Ramp, building agentic procurement for 70,000+ businesses.
I'm extremely grateful to say that the things I've built have been collectively used by and impacted 5M+ people. Previously, I was one of 2 engineers building the AI Qual product from the ground up at Pogo, serving customers like Uber and Point72. In high school, I developed AI solutions for the Indian government in reducing air pollution, currently used across 15+ cities to help policymakers make data-driven decisions on air pollution.
I lead recruitment at USC's premier startup accelerator and ran USC's flagship incubator, helping launch 14 companies out of USC's startup ecosystem.
Work
Cool Stuff I've Built
Relay
Run and manage parallel computer use agents locally within containerized sub-environments, view workstream recordings, and hop in to take over the desktop for true human-agent collaboration.
Cortex
A secure multi-agent reasoning and action engine. Cortex launches multiple task-based agents in parallel dynamically generated by a master planning agent.
Intervene
OS level AI agent that takes over and completes tasks for you. Tell Intervene what to do, and it will do it for you, autonomously in the background.
NightShift
Autonomous background coding agent that makes full-stack software changes in isolated sandboxes, built using Modal and Opencode. Give Nightshift a task, and it will execute/test the code, verify the UI, and draft a pull request for you in the background.
IdeaVine
Multi-modal mindmapping tool designed to help you think freely. Let AI think with you, not for you.
Hera
Air Pollution analytics tool that provides the best solution to the air pollution at the neighborhood level to policymakers.
Tone
Wear your wisdom. Tone captures your lived experiences, your conversations, your ideas, your to-dos, automatically.
AIlytics
AI data scientist that connects directly to your databases and derives insights and visualizations through simple natural language prompts.