Portfolio · Miami, FL

Ezra Pinsky. Full stack developer.

Five years and counting of shipping production software end to end: data models, backends and APIs, React frontends, third party integrations, and the infrastructure it all runs on. My recent work centers on AI document processing and case management platforms that I own from first scoping call through production.

Open to full time roles. Resume on request: ezpinsky@gmail.com

5+
years shipping production software
780+
tests in the Appeals Doctor pytest suite
35%
Lighthouse performance gain at Via.work
9
domain apps in production, one engineer
Featured build · Sep 2024 to present

Appeals Doctor

An AI case management platform that automates reimbursement appeals for Amazon sellers. I am the sole engineer: I designed the system, built it, and operate it in production.

  • Django 5, PostgreSQL, and Celery (Redis locally, Amazon SQS in production), organized into 9 domain apps. I own the data model and its schema migrations.
  • Built a Google Gemini pipeline that validates invoices and supporting documents before a case is ever submitted.
  • Took over a legacy OpenAI drafting service and rebuilt it into a database-configurable prompt system that writes appeal letters from case data.
  • Document intake built for messy reality: S3 storage under GUID keys, PDF structure validation with pypdf, and Tesseract OCR with a cloud fallback for scans.
  • Amazon SP-API, Salesforce case sync, Stripe billing, Gmail ingestion, and SendGrid integrated into one resilient, idempotent case-processing pipeline.
  • 780+ pytest tests with coverage reporting, GitHub Actions CI/CD with Ansible, systemd, and Nginx, and per-environment Sentry monitoring.
Django 5PostgreSQLCeleryGeminiOpenAIAWS S3 + SQSStripeSalesforcepytest
appealsdoctor.com
Case pipelineLive in production
01

Intake

Documents arrive by upload or Gmail ingestion, stored in S3 and validated with pypdf. Scans go through OCR.

S3 · pypdf · Tesseract
02

Validate

A Gemini pipeline checks invoices and supporting documents before the case moves forward.

Google Gemini
03

Draft

An OpenAI service writes the appeal letter from case data, driven by prompts configured in the database.

OpenAI · prompt system
04

Submit and sync

Cases flow through Amazon SP-API with Salesforce sync, Stripe billing, and SendGrid updates along the way.

SP-API · Salesforce · Stripe
Selected work

More things I've shipped.

SaaS · Insurance tech

Claim Compass

An insurance claims management SaaS. I contribute production features across the stack: Next.js 14 App Router screens built with shadcn/ui and Radix, form flows validated with Zod and React Hook Form, Supabase auth, Stripe billing, and AI document processing with Gemini and OpenAI.

Next.js 14TypeScriptTailwindSupabaseStripe
claimcompass.io
Fintech · Lending

Park Business Capital

A customer portal for a merchant cash advance brokerage. React 18 with TypeScript on Vite, backed by a serverless Node.js API on AWS Lambda and API Gateway with S3 presigned uploads, DynamoDB, and JWT auth. A second Express and PostgreSQL backend adds Socket.io real-time features and a fault-tolerant Salesforce integration with OAuth 2.0 and retry with backoff.

React 18AWS LambdaExpressPostgreSQLSalesforce
parkbusinesscapital.com
HR tech · Acquired by Justworks

Via.work rebuild

Rebuilt the marketing and content frontend in Gatsby, React, and TypeScript for a 35% Lighthouse performance gain. Migrated several hundred components from Strapi to Contentful and optimized GraphQL queries so content teams across multiple regions could work faster.

GatsbyTypeScriptContentfulGraphQL
Earlier and elsewhere

The rest of the story.

Reimbursement Genie2020 to 2022 · Lead developer

Amazon reimbursement recovery software in C#, ASP.NET MVC, and SQL Server. Built the SP-API integration layer (LWA auth, AWS STS, Signature V4), led a company-wide TDD initiative with MS Test, and ran agile sprints on Azure DevOps. The features I shipped cut manual labor by 40% and grew revenue by 15%.

EZ Web clients2021 to present · Independent practice

A Shopify storefront and ad-spend analytics dashboard for an e-commerce brand, a serverless DeepL translation system for Eleveo that saves the team 10 to 15 hours of manual work weekly, WordPress rebuilds for the Dolan Law Firm and others, and Python utilities for OCR, image optimization, and email automation.

Open sourceAmazon SP-API community

Proposed API usability and documentation improvements and helped other developers resolve authentication and integration issues.

Stack

Tools I reach for.

Comfortable across ecosystems: Python and Django, Node and TypeScript, React and Next.js, C# and .NET.

Frontend
ReactNext.js (App Router)TypeScriptGatsbyViteTailwind CSSshadcn/uiTanStack QueryZustandReact Hook FormZod
Backend
DjangoNode.jsExpressFastAPIFlaskASP.NET MVCCeleryREST API designWebSockets
Data and cloud
PostgreSQLSQL ServerDynamoDBRedisSupabaseAWS (Lambda, S3, SQS, API Gateway)DockerGitHub ActionsAnsibleNginxSentry
AI and integrations
OpenAIGoogle GeminiTesseract OCRPyMuPDFLLM evaluationStripeSalesforceAmazon SP-APIShopify Admin APISendGridContentful
Ezra Pinsky
About

I like owning the whole problem.

I came to software through App Academy after building my first systems in Excel and VBA at a precious metals refinery, where my chain-of-custody work earned a respected third-party certification. Before that I founded and ran a media digitization business.

That path shaped how I work now. I am happiest when I own a product end to end: talking to the client, designing the data model, shipping the frontend, and carrying a pager for the result. Appeals Doctor is what that looks like in practice.

I live in Miami, FL, and I am currently open to full time software engineering roles.