Abhigyan (Abhi) Arya is an engineer at Reducto and the former cofounder and CTO of Opennote. His work spans document infrastructure, search, real-time systems, AI agents and human-computer interaction.[1]
Start with Opennote acquisition announcement →
GitHub 
X · abhiarya Career
Arya cofounded Opennote, an education startup in Y Combinator’s Summer 2025 batch. Reducto acquired the company in 2026. In his acquisition announcement, Arya describes the connection between the two businesses: turning large collections of notes, documents and scattered information into material people can use. His current personal site lists product, software, infrastructure and applied AI work at Reducto.[1][2]
Contributions
Arya’s retained résumé describes responsibility for Opennote’s backend infrastructure, multi-agent architecture, real-time collaboration, developer APIs and text-to-video pipeline. That scope connects the visible learning experience to retrieval, security and delivery systems. His personal site also lists work as Browserbase’s first engineer on Director and Gateway flight-control and mission-operations software at NASA Johnson Space Center.[3][1]
Ideas
Arya’s acquisition post frames the learning problem as making information understandable and usable. Opennote aimed to help students practice and retain material, going beyond storing a document or returning an answer. This explains his continuing interest in document and search infrastructure: extracting information is one stage, while an interface must turn it into something a person can act on.[2]
Tools, writing & talks
Writing, research & talks
Start here
Opennote acquisition announcementlinkedin.comArya explains the learning problem behind Opennote and why document understanding connects it to Reducto.
Read his acquisition statement first for the product problem that connects his education startup to document infrastructure. It needs no technical prerequisites and explains the motivation behind the engineering work.
Author[2]Reducto's Abhi Arya explains why a one-tool-per-endpoint MCP server broke on day one and how a rebuild with fewer, structured tools, a state snapshot tool, confidence scores and usage instrumentation produced an agent the team trusted.
Catalog workOriginal source