Sean Keith
15+ years building and scaling full-stack, serverless, and AI-driven systems. Built KERNSTACK's distributed core, console, CLI, and SDKs, agent-first. B.Sc. Business Informatics, B.A. Audio Engineering.
Investors · Seed round · Closed preview
Software is now built by people and agents together, and intelligence has become a commodity. What is scarce is a safe environment where it can act on real systems. KERNSTACK is that environment: conversation, intelligence, code, and infrastructure as one system, on bare metal we operate — with intelligence plugged in from outside and control kept inside.
Teams prototype in minutes; production still takes months, because one product on a hyperscaler means a dozen services with separate APIs, IAM models, quotas, and invoices. That collage just got two new kinds of user: coding agents that inherit whatever cloud you hand them, and clients and partners who need to see and steer the product without console access.
Not retrofitted from VM primitives. Not resold hyperscaler capacity. Not a tool beside the cloud. Serverless-first, one control plane, built for owned hardware with User Nodes on-premise, and built from day one for people and agents working in the same environment.
Agents are the new infrastructure consumers, and frontier models are interchangeable through harnesses. The scarce asset is the environment that gives them context, permissions, credits, and a ceiling — and is the system of record when the model changes.
Four layers as one system. Collaboration: conversation-based product development via console, Telegram, or WhatsApp. Intelligence: agent workers and sandboxes running pluggable harnesses plus hosted and AutoML models. Code: the whole application as one stack, deployed with one command. Infrastructure: compute, data, search, inference, and training on one control plane, on hardware we operate.
Subscriptions with monthly credits, credit packs on demand, enterprise agreements with SLAs and dedicated zones. One meter across compute, storage, search, training, inference, and agent runs — agent compute is a revenue line, not a cost centre. Owning the stack from hardware to API makes pay-for-what-runs a margin decision, not a subsidy.
Partners see the product before they see the platform. A services unit builds and migrates client applications onto KERNSTACK for immediate usage and feedback; every project invites clients and vendors, each a future builder. Blueprints, SDKs, and the Cursor plugin handle zero-touch developer acquisition; platform credibility opens regulated sectors top-down.
Use of funds: 51% engineering including agent compute, 20% infrastructure — three own arm64 clusters, one metro at a time, starting across three Berlin data centers — 8% go-to-market, 4% compliance and security, 5% operations, 12% reserve. First hires: a platform engineer (Go, distributed systems), an applied ML engineer (AutoML, training pipelines), and a second platform engineer with a DevRel profile (SDKs, blueprints, builders).
15+ years building and scaling full-stack, serverless, and AI-driven systems. Built KERNSTACK's distributed core, console, CLI, and SDKs, agent-first. B.Sc. Business Informatics, B.A. Audio Engineering.
Electrical Engineering PhD, M.Sc. Computer Science. 15+ years in analog IC design at TI, Intel, IDT/ZMDI, and Renesas. Secure, ISO-aligned infrastructures, incident analysis, and system hardening; cybersecurity lead for regulated customers.
Deck, data room, and a demo of the closed preview on request.