Lucas Rangel

Summary

Senior Full Stack Software Engineer specializing in early-stage startups and AI/ML systems.

Architected LangGraph-based agent workflows for contract drafting and review with retrieval-augmented generation for contextual document understanding.
Built evaluation framework measuring agent goal and tool call accuracy.
Built observability with OpenTelemetry for logging, metrics, and alerting in production.
Built tool execution framework with state synchronization enabling simultaneous AI and manual document editing.
Built Next.js application with real-time document collaboration UI.
Developed type-safe database layer using Drizzle ORM with multi-tenant data isolation patterns and SOC 2 compliance.
Architected GCP infrastructure with Terraform, including Cloud Run and Pub/Sub message queuing.
Built bidirectional data transformation layer between Slate editor JSON format and Word Docx, preserving formatting and metadata.

As first engineering hire, built product from concept to hundreds of clients with thousands of daily users, owning full technical architecture and implementation using Ruby on Rails, Hotwire, Stimulus Reflex, Phlex, and Sidekiq.
Built production observability, error monitoring, and alerting systems.
Shipped features weekly with automated testing, crucial for rapid iteration during product discovery phase.
Built AI chatbot for real estate lead qualification, increasing client closing rates by 30% through automated conversation flows and follow-up.
Drove product direction and roadmap, influencing key decisions and features based on customer feedback and market validation.
Managed infrastructure across Azure and AWS using Terraform, Docker, and Kubernetes.
Set up CI/CD pipeline for development and deployment processes.

Implemented new features end to end, utilizing Node.js, React, React Native and PostgreSQL for full-stack web and mobile applications.
Developed internal tools using Rust to enhance developer productivity.
Built code mods to aid in large refactoring projects, reducing technical debt while saving hundreds of engineering hours.
Mentored junior engineers on best practices, code quality, and technical problem-solving.
Solved Node.js performance and scaling issues by fixing memory leaks and improving memory usage, reducing startup time by ~2x and memory usage by ~4x.

Developed a credit score model using machine learning algorithms to assess customer debt repayment likelihood.
Built Hadoop data pipelines using Apache Airflow, Hive, Python, and Scala.
Worked with query optimization for large datasets including geospatial data.

Languages: Ruby, Rust, TypeScript, Python, SQL, JavaScript
Technologies: React, Node.js, PostgreSQL, Redis, Docker, Kubernetes, LangGraph, Next.js, Ruby on Rails
Concepts: AI/ML Systems, Distributed Systems, Machine Learning, DevOps, CI/CD, Observability
Tools: Git, Terraform, AWS, GCP, Azure, OpenTelemetry

2014--2019

—

a Lisp interpreter in Rust.

—

a Language Server Protocol implementation in Rust for Phlex templates.

—

a DOCX-to-Plate deserializer written in Gleam.

—

a TypeScript framework for building MCP servers.

Decorator-based tool registration with automatic Zod schema generation from TypeScript types
Toolkit system for grouping related tools with namespacing to avoid context bloating
Published on npm with production-ready compliance test suite

—

an MCP server bridging Obsidian vaults with AI assistants.

Multiple tools for note management, task tracking, and flashcard generation
Vault-based prompts and resources: Users can store and edit their prompts and resources in Obsidian and use them in any MCP client

—

a type-safe configuration tool for Karabiner-Elements using Apple's Pkl.

Declarative DSL for defining keymaps with autocomplete, type safety, validation, and editor support
First-class support for popular macOS window managers and applications, live-reload for instant configuration updates
Published as a Homebrew package