We are building the secure network where your AI agent talks to other people's agents, so humans stop doing the repeated coordination. The longer vision is full Agent-to-Agent handshakes that 10x communication efficiency. To start: instead of sending a static document, you send a link where people can talk to your AI to get answers and book meetings. That's "limited agent deployment": zero signup for recipients, and the first real deployment of our cross-boundary delegation protocol.
We are addressing the Coordination Tax, where knowledge workers spend over 50% of their time on coordination rather than execution. Current AI assistants help with drafting and summarizing but can't safely interact externally due to risks of hallucination and information leaks, creating the Security Paradox: more context makes AI more useful but also more dangerous. Pulse solves this with access-aware delegation, enabling AI to assist communication without breaching information boundaries.
Competitors: AI social apps (Elys, Second.me)
What they do: Have AI learn your tone to social with others.
Why that’s not enough: Real external coordination needs personalized context (calendar, real data). That hits the security paradox: giving an agent full context for personalization creates serious risk. Current agents (OpenClaw, Manus) are built for single-user, isolated use, not for “someone talks to your AI via a link.” In fact, most startups avoid confronting the security paradox and have to choose alternatives, but time will show that the bottleneck has to be fixed.
Pulse: Access-aware from day one. Agents run in isolated context cells & smart protocols and only see what’s explicitly allowed. That enables safe cross-boundary coordination.
Research: I co-drafted a white paper with Stanford, Oxford & Columbia. Built the first benchmark for cross-boundary collaboration under distributed access (10k QA pairs) and an agentic access protocol that 2x SOTA. Backed by Torr Vision Group, Oxford.
Product: 300+ iterations to build a robust agent with personalized context. After pivoting, I ran 50+ interviews to validate demand and onboarded 10 users who share their Pulse link 10+ times per day.
TAM ≈ $100B+ — The entire global enterprise collaboration, communication, and productivity layer (the "Death of Sync" market).
SAM ≈ $15B — External collaboration, secure data rooms, and scheduling automation (DocSend + Calendly + Intercom).
SOM (realistic 3y) ≈ $10M–25M ARR — High-leverage external communications for founders, VCs, and early-stage tech teams
I am a systems and AI builder. Oxford Engineering + Incoming Columbia CS PhD. I won the 2025 NeurIPS Best Poster award (AI for Finance) and have founded an open-source agent framework with over 10k+ stars. I’ve shipped 300+ iterations in months and built a 500-person waitlist of founders using Pulse to automate their workflows. Previously, I researched agentic intelligence at Meta and Microsoft Research, and founded WeLight (an AI platform with 2,000+ users).
I built an AI system to automate my entire job hunt—resume tailoring, cover letters, outbound emails—and used it to land Meta and multiple top offers. Then I turned it into a product for 2000 users.
I flew myself to Silicon Valley to meet founders building agent systems because I believe networked agents will reshape how humans work. I turned down FAANG & big offer to go all-in on building that future.
Technical: At 14, I wrote an AI textbook and founded my high school’s first AI society by self-studying Calculus.
Non-Technical: I left home and began living alone at 12. It taught me resilience, self-discipline, and how to build a life from scratch. That independence shaped how I lead, learn, and build today.
24 hours before a NeurIPS workshop deadline, a YouTube Short explaining Reynolds’ Boids rules sparked an idea about emergent coordination in AI agents. I spent the next day building a prototype and writing ~20k lines of code, turning the idea into a full academic paper submitted just before the deadline. It was accepted and later helped me get a Columbia CS PhD offer. I think this reflects my habit of rapidly connecting ideas and executing at high speed.