Projects

Three products. One engineer. All open source.

I don't have a portfolio of mockups — I have systems that other people run. Each of these is public code you can read, clone, and break.

Founder · architect · operator
Apache-2.0 · Go

A cloud platform where every workload runs inside its own Firecracker microVM. I designed and built the whole stack: the host agent that orchestrates VMs, a snapshot-restore boot path that gets a machine serving in 179ms, a pre-allocated network-namespace pool, copy-on-write VM forking, managed PostgreSQL with point-in-time restore, git-driven app hosting with blue-green deploys, serverless functions, two SDKs, and usage billing.

179msp50 boot, measured in prod
~400Go files across 20 packages
300+organizations signed up
16,384sandbox ceiling per host

The hard parts: userfaultfd demand-paging VM memory from object storage, XFS reflink for copy-on-write disks, hibernate/wake for scale-to-zero, and a multi-region scheduler with lease-based ownership.

GoLinuxGoogle CloudTerraformAnsiblePostgreSQLClickHousePrometheus
Creator

Riff

Open source · TypeScript

Prompt → a running full-stack app, then fork it. A self-hostable alternative to Lovable / v0 / bolt where every generated app runs in a real microVM instead of a browser sandbox — so it can have a database, a real server, and a URL that stays alive. The differentiator is branching: fork a running app and its database to explore several design directions in parallel, then keep the winner.

3 frameworksVite/React, Next.js, static
Fork + DB clonebranch an app and its data
$0 idleapps hibernate, wake on visit
Self-hostedyour keys, your infra
TypeScriptNext.jsPostgreSQLDockerLinux
Creator

PandaFlow

MIT · TypeScript

A visual builder for AI agent workflows. Drag-and-drop LLMs, databases, APIs and code-execution nodes onto a canvas, wire them together, and deploy the result. Code nodes execute in isolated microVMs, so a workflow can run untrusted or generated code without putting the orchestrator at risk.

Visual canvasReact Flow node editor
Isolated execcode nodes run in microVMs
MITfork it, ship it, sell it
TypeScriptNext.jsDockerPostgreSQL
Why hire me

I ship whole systems, not tickets.

Fourteen years of infrastructure work, and the last stretch spent proving I can take something from an empty repo to a platform other people depend on.

01

Range most engineers don't have

Kernel-level isolation and Terraform modules and the billing integration and the docs. I've had to own every layer, so I know where they actually break each other.

02

I've been the pager

Solo on-call for a multi-tenant platform teaches you to design for the bad Tuesday: blast radius, recovery paths, and alerts that fire before a customer notices.

03

Speed you can verify

A complete cloud platform in about six months, alone. The commit history is public — you don't have to take my word for the pace.

04

I care about the bill

I built scale-to-zero and active-CPU billing because I pay my own infrastructure invoice. Cost is an engineering constraint to me, not a finance problem.

Services & rates →hello@ajayk.sh