Avi Flombaum

Founder & Product Engineer

GitHub Resume

Hi Shopify! You asked to see things I've built that are being used.

These products show how I ship fast, solve real problems, and use AI to amplify what's possible.

Scroll to explore

Current Projects

CTO/EIR @ Innovent Capital and Personal

PromptEngine

Open-source Rails engine for managing AI prompts at scale

Live Demo GitHub
PromptEngine testimonial

The Problem

Everyone's hardcoding prompts in their applications. But prompts are data and data doesn't belong in code. Data needs versioning, testing, and management.

What I Built

A Rails engine that treats prompts as first-class citizens. Automatic variable detection, composable workflows and tool calls, version control, logging and telemetry, and A/B testing and evals built in.

Currently Using It For

Powers all the AI features in my current projects, including the VC matching system that helps startups find the right capital.

Smart Prompt Builder
Smart Prompt Builder
Composable Workflows
Composable Workflows
Production Use Case
Production Use Case

VCMatch.ai

AI-powered VC discovery that matches your startup with investors who actually fund companies like yours. No spray-and-pray, no cold outreach to the wrong partners. Built on PromptEngine's adaptive prompt infrastructure.

Demo Live Beta

Socratic Tutor

Learn anything through guided questions, not answers. An AI tutor that adapts to your knowledge gaps and teaches you to think critically. Powered by PromptEngine's conversational prompt framework.

Try It

My AI Workflow Standardization

You asked about standardizing AI workflows for developers. Here's what I've done:

RSpec Rails Agents

12 specialized AI agents that write production-ready Rails tests

RSpec Agents
Rails Master Agent
Rails Master Agent
Model Specs
Model Specs
Request Specs
Request Specs
System Specs
System Specs
Isolation Expert
Isolation Expert
ActionCable Specs
ActionCable Specs
ActionMailer Specs
ActionMailer Specs
ActiveStorage Specs
ActiveStorage Specs
ActiveJob Specs
ActiveJob Specs
Fixture Expert
Fixture Expert
DRY Principles
DRY Principles
Helper Specs
TDD Advice

You wouldn't ship software without CI/CD, testing frameworks, or code review. The same discipline is now required for AI. I built a framework that gives teams a shared system to create, evaluate, and continuously improve agents. It standardizes how agents are designed, measured, and deployed, so organizations can move from one-off experiments to reliable, repeatable workflows.

What I Built

A framework of specialized agents that transform AI assistants into expert test writers. Each agent has deep domain knowledge. Model specs know about validations and scopes, Request specs understand authentication and APIs, Isolation agents handle mocking external services.

How It Works

Instead of writing prompts, developers reference agent files. Works with Claude Code, Cursor, Copilot, Codex, etc. Apply globally or per-project. The standardization pattern is reusable, tested, evolving agent definitions.

Compounding AI

Imagine every team at Shopify working with the same family of AI powered workflows that are observable, testable, improvable, just like people. A code review workflow can be A/B tested, graded on defect rates, and given a performance review, then promoted, retrained, or retired. Incident workflows can be measured on time to resolution and blast radius containment, then tuned globally when one team finds a better pattern. Compounding engineering turns process into a first class artifact that improves with every deployment, which means coordination at scale becomes a feature of the system, not a tax on the people.

Where I Can Help Shopify Today

Developer Tools

Internal AI workflows that help teams ship faster and safer

Standardized AI Agents

Turn Sidekick-style tools into reliable, reusable infrastructure

Education Platforms

Help merchants and developers learn, adopt, and experiment

Flatiron School

Founder, Dean, and Chief Product and Technology Officer

Learn.co

The LMS that became GitHub Codespaces

4000+ Curriculum Commits Platform Deep Dive
10,000+ Students
$14M Exit to 2U
60 Team Size

The Problem

Traditional coding education was broken. Students watched videos and typed in sandboxes, never learning real development workflows.

What I Built

A learning platform built on top of GitHub that treats every lesson as a repository. Students fork lessons, write code to pass tests, and submit pull requests. The IDE launches in-browser with everything configured in under 2 seconds.

The Result

We taught 10,000+ students to code with a 90%+ satisfaction rate. The GitHub team was so impressed they hired half my engineering team to build Codespaces. We sold the platform to 2U for $14M.

One-Click IDE Launch & Automated Testing
One-Click IDE Launch & Automated Testing
Real Git Workflow: Fork → Code → Submit PR
Real Git Workflow: Fork → Code → Submit PR

Revature

Chief Product Officer (2020-2023)

eXa Learning Infrastructure

Enterprise training platform at scale

2000+ Contractors Deployed
Enterprise Fortune 100 Training

The Scope

Led product organization building interconnected learning systems for enterprise technical training. Architected and delivered the full product ecosystem powering distributed instruction teams and thousands of concurrent students.

The Approach

Built modular platform architecture treating each stage of the learner journey as a distinct but connected system—from pre-training assessment through content delivery, real-time instruction, and quality monitoring.

Course Platform & Learning Portal
Course Platform & Learning Portal
Integrated Code Editor & Lesson Interface
Integrated Code Editor & Lesson Interface

Product Ecosystem Delivered

eXa LMS

Core learning platform & content delivery

Quality Monitoring

Efficacy dashboards & program analytics

Distributed Teams

Instructor tools & team coordination

Executive Impact

The Pattern: Second learning platform at scale. From founder-built (Flatiron) to executive-led product organization (Revature). Same core insight—real development workflows, not sandboxes—applied to enterprise training infrastructure.

Rails Ecosystem & Community Leadership

Rails community contributions and education impact at scale.

shadcn/ui for Rails
840+ stars

shadcn/ui for Rails
NYC on Rails
3,900+ members

NYC on Rails
Kode with Klossy
10,000+ scholars

Kode with Klossy
My Work Log My Code Blog

Let's Talk

I'd love to learn about what Shopify is actually working on and explore where I might be helpful. Whether it's developer tools, AI workflow standardization, or something completely different - I'm excited to dig in.

Email me
GitHub Code Blog Resume