John Esten Davenport

Fuel card verification platform with 5-validator fraud detection pipeline, Claude Vision OCR, LLM anomaly analysis, Stripe Issuing/Treasury/Connect, and Twilio SMS workflows. Built with CodeMySpec — 55 commits, 5 days, zero human-written code.

Marketing analytics platform correlating ad spend across Google Ads, Facebook Ads, GA4 with QuickBooks revenue. Pearson + time-lagged cross-correlation, Claude-powered insights, Vega-Lite dashboards. Built with CodeMySpec — 40 commits, 13 days, zero human-written code.

Published Hex package with distributed test runner (file-based locking, event caching, result replay) and JSON ExUnit formatter. Powers CodeMySpec's Claude Code extension.

LLM-driven marketing operations for CodeMySpec using MCP integrations (GA4, Reddit, Search Console, SEO tools). Dictation-to-cleanup workflow for authentic content. Getting reliable signups.

Claude Code extension orchestrating the full Elixir/Phoenix/LiveView development lifecycle: requirements, architecture, BDD specs, code generation, QA. 368 modules, 46 contexts. Two production apps built with it.

Natural language to dashboard tool using Snowflake Cortex API for SQL/Vega-Lite generation and Claude for layout orchestration. Dynamic Blazor component rendering from LLM-generated JSON. MVP with static visualizations.

Multi-environment Python/SQL framework for Snowflake with config-driven data mart deployment, projector dependency chains, and PowerBI/Fabric credential automation

Used LLMs to validate 170+ report migrations from Tennacle to Snowflake by iterating on queries until results matched production baselines

TypeScript/Node.js application aggregating marketing metrics across Google and Facebook platforms, with multi-tenant architecture, OAuth2 integration, and data normalization pipeline.

Middleware that lets AI models interact with REST APIs by generating MCP tools from OpenAPI specs. .NET with DDD, dual-database (SQLite/PostgreSQL), OAuth2/API key/Basic auth.

Full-stack application for multi-tenant Git repository management with browser-based operations using Blazor WebAssembly and ASP.NET Core. Implements clean architecture patterns and domain-driven design.

Final iteration of the AI-assisted development experiment series, testing C# and .NET as a better-suited language for LLM collaboration, resulting in a multi-tenant todo app with real-time updates.

Python-based code generation tool for Rust/Axum applications that automates creation of models, repositories, and services using template pattern matching and LLM-assisted code analysis

Third iteration testing an LLM-driven documentation synthesis approach to project bootstrapping, building a Rust/Axum todo app with hexagonal architecture.

Command-line tool to automate context preparation and task management for AI pair programming sessions, analyzing codebases to gather relevant documentation and code for LLM assistance.

Fourth iteration exploring LLM-assisted development in Rust, testing the boundaries of AI code generation for repetitive patterns and discovering where traditional tooling works better.

Second iteration exploring AI-assisted development in Rust, focusing on documentation-first approaches to LLM collaboration while building a layered todo application with JWT auth and rate limiting.

Built a Rust todo application as a controlled experiment in AI-assisted development, documenting prompting strategies and context management techniques while learning Axum, Sea-ORM, and Rust patterns.

Desktop application for real-time audio transcription and entity recognition using WhisperX and SpaCy, built as an AI-assisted learning project.

Full-stack platform that automates influencer discovery and content analysis using vector similarity search and semantic embeddings, built with FastAPI, React, and PostgreSQL/pgvector.

Spike for a real-time meeting transcription application implementing voice activity detection, speaker diarization, and speech-to-text using Elixir and ML models, with native desktop integration.

Contractor matching platform using Phoenix LiveView, featuring real-time updates and scoring algorithms for connecting ERP professionals with project opportunities

Conversation intelligence system processing communications using AI pipelines, real-time transcription, and semantic analysis capabilities

Complete port of Pixelmatch from JavaScript to Elixir, implementing image comparison algorithms with antialiasing detection and color science optimization

Built a web automation platform using Elixir/Phoenix that enables teams to automate documentation screenshots with real-time collaboration and visual regression testing

Elixir/Phoenix platform that automates visual documentation through browser control, real-time collaboration, and element tracking

Built an automation platform that transforms manual screenshot collection into an automated pipeline, featuring distributed architecture across web, desktop, and browser extension components with annotation and session management capabilities

Built a documentation automation system using Elixir/Phoenix, featuring real-time collaboration, browser automation, and distributed state management across OTP applications.

Engineered a domain-driven automation system that transformed complex Manufacturing Execution System implementation planning from manual spreadsheets into a streamlined pipeline, reducing project estimation time from days to hours while improving accuracy.

Designed and implemented a comprehensive Technical Writing system handling complex manufacturing execution concepts, with information architecture and integration pattern documentation.

High-performance async Python server enabling real-time industrial equipment monitoring through modern web interfaces, handling concurrent data points with low latency

Modern Vue.js frontend for manufacturing execution systems featuring real-time WebSocket data streaming, recursive component architecture, and reliable operation

Django-based contact management system with automated profile enrichment via Pipl API, featuring probabilistic matching and granular permissions for sensitive data protection

Python framework for integrating manufacturing equipment using OPC UA, enabling real-time monitoring and process control across multiple facilities

Architected and deployed a modern manufacturing integration platform that eliminates custom backend development by combining real-time PLC data streams, legacy database synchronization, and automatic API generation