Mosaia
Mosaia is an AI agent platform that lets businesses and users create, share, and run autonomous agents in the cloud — with first-class integrations into the apps where work already happens, like Slack and Telegram. Co-founded and led as CTO.
- Date
- January 2024 – May 2026
- Role
- Co-Founder & CTO
- Tags
- AI Agents, LLM Platform, SaaS, Integration Platform, Cloud Architecture, Slack, Telegram, B2B
Mosaia is an AI agent platform that lets businesses and individuals create, share, and run multiple AI agents in the cloud, with first-class integrations into the third-party tools where work actually happens. As co-founder and CTO, Aaron architected, developed, and shipped the first four versions of the platform, built and led the engineering team, and contributed to fundraising, go-to-market, and growth strategy.
v1 — Integrating AI Into the Apps People Already Use
The first iteration of Mosaia was an integration layer that let AI and LLM-powered behaviors be embedded directly into the third-party applications businesses already lived in — most prominently Telegram and Slack. Rather than asking teams to come to a new surface, Mosaia met them where they were, exposing LLM reasoning and structured workflows through the chat tools their teams already used every day.
v2 — AI-Powered RFP Web App for B2B Procurement
The second product was a front-end, AI-powered web application focused on B2B sales workflows — specifically, turning the painful, document-heavy RFP (Request for Proposal) process into something teams could spin up, iterate on, and respond to quickly. The application combined LLM-driven drafting and analysis with a workflow designed around the realities of enterprise procurement.
v3 — Multi-Agent Cloud Platform
The third version generalized the lessons from v1 and v2 into a platform: a cloud-hosted environment where businesses and end users can create, share, and run multiple AI agents, each composed of prompts, tools, memories, and third-party integrations. Agents can collaborate, be embedded into external surfaces, and be invoked across the integration channels built in v1 — bringing the "agents in the apps you use" thesis full-circle on top of a generalizable agent runtime.
v4 — DAG Orchestration & RAG Drive
The fourth version deepened the runtime into a true coordination layer: a plan-task orchestration engine that decomposes work and executes it across a DAG of LLM agents, so multi-step tasks flow through dependent stages rather than a single monolithic prompt. Backing it is an AI-powered cloud drive that grounds agents in a managed RAG context — documents and shared state are indexed and retrieved on demand, giving every agent in the graph the knowledge it needs at the right step.
Technologies & Architecture
- LLM orchestration and integration across multiple model providers
- Cloud-native, scalable architecture designed to host and run agents on demand
- Integration adapters for Telegram, Slack, and other third-party surfaces
- Multi-agent runtime supporting agent composition, tool use, and shared resources
- Plan-task orchestration engine executing work across a DAG of LLM agents
- AI-powered cloud drive providing managed RAG context and retrieval for agents
Team & Leadership
Built and nurtured the technical engineering team that shipped and maintained the platform. Set the architectural direction across the first four versions, managed the technical resource and expense plan around the product build, and partnered closely on go-to-market and fundraising as a founding executive.