Joshua Darron Phillips

Hi, I'm Joshua

Founding engineer at RocketRide.

I build the developer-facing surface: SDKs, MCP server, VS Code extension, workshop tooling.

View Projects
MySQL MongoDB TypeScript Python Go Vim Ollama OpenAI Claude

Founding-team engineering across the developer surface.

SDK & Tooling

TypeScript, Python, and Node.js SDKs, MCP servers, and VS Code extensions for an open-source AI runtime.

Full-Stack Engineering

End-to-end product work in TypeScript and Python: Postgres, Neo4j, Next.js, and the runtime layer underneath.

Data & AI Systems

AI agents, pipeline architecture, and graph data science. Production systems, not demos.

Backend Systems

Data-intensive services, streaming APIs, and the runtime layer underneath the orchestration frameworks.

Joshua Darron Phillips

Founding engineer from the Golden State

I'm a founding-team engineer at RocketRide, an open-source MIT-licensed AI runtime. I own the developer-facing surface: the TypeScript and Python SDKs, the MCP server, the VS Code extension, and the workshop tooling.

I also built Pulsar, a market intelligence agent running internally across finance, HR, and marketing at RocketRide, which launches as a featured app on RocketRide Cloud. I work in TypeScript, Python, and C++, and I write about what I build.

Generative AI

Observability for AI Pipelines: What to Trace and Why.

Most teams treat AI observability as a retrofit. It's a design constraint.

Level Up Coding

MCP Is the Connective Tissue Your Agent Stack Is Missing.

A protocol-level explanation of why Model Context Protocol matters more than its mention volume suggests.

Level Up Coding

The Palimpsest Library.

Rebuilding documentation architecture around two first-class audiences: humans and the coding agents that work alongside them.

Level Up Coding

Two Communities Are Building on MCP. They Think It Is Two Different Things.

The protocol split between agent tooling and data-integration camps, and why each side reads the same spec differently.

Level Up Coding

The Full Stack Is One Layer Deeper.

Argument for naming and formalizing the AI layer of the stack the way LAMP, MEAN, and JAMstack named earlier patterns.

Level Up Coding

I Tried to Use My Own Product for Everything.

Postmortem on three architectural wrong turns building Pulsar, and the principle I should have started with.

Generative AI

I Expected Pulsar to Land on the Repo Shelf. It Didn't.

How a weekend dogfood project ended up adopted by finance, HR, and marketing.

Why Your RAG Pipeline Is Slower Than It Needs to Be.

Most "local LLM is broken" complaints are pipeline problems, not model problems.

Your Pipeline Is Already an MCP Tool.

How existing pipeline architecture maps directly onto the Model Context Protocol surface.

Your IDE Already Runs Your Code. Now It Runs Your AI.

The IDE is becoming the runtime for AI-assisted work, not just the editor for it.

Pulsar Told Us We Were Wrong About Go.

A finding from running Pulsar against real developer-source data that contradicted what the team had assumed about Go.

Tell me about your
next project

Email: phillips@joshuadarron.com · Writing: Medium