rafael.medeiros
PTEN
Resume

~ $ whoami

RafaelMedeiros

Software engineer focused on back-end.

I build and maintain systems that serve customers every day, with Ruby on Rails, TypeScript and PostgreSQL. My focus is architecture, performance and reliability. I'm studying Computer Science at Feevale.

What I work with

I'm a curious and dedicated Software Engineer with solid experience in systems development and automation. I focus on building scalable, reusable back-end solutions with Ruby on Rails, JavaScript, Java and Python, always applying clean, solid architecture and automation practices. I value high-quality code, teamwork and continuous learning.

LanguagesRubyTypeScriptJavaScriptPythonJava
Back-endRuby on RailsSidekiqEdge Functions (Deno)REST APIsWebhooks
DataPostgreSQLRedisSupabaseRLS
Infra and qualityDockerCI/CDAWS LambdaPlaywrightGit and GitHub
Automation and AIn8nTypebotPrompts and LLMsClaude Code
stack01

What I've built

At New Byte, a startup that builds automation for short-term rentals, and in my personal projects. Pick a file to see what I delivered.

team_authorization.rb

NChat

Team-based authorization

I created a single authorization rule in Ruby on Rails that decides what each agent can see, reused across lists, counters, filters, the kanban board and access by link. As a result, each team sees only its own conversations, operations become more secure and better organized, and nothing changes for customers who don't turn the option on.

Ruby on RailsPostgreSQL
link_preview_job.rb

NChat

Resilient link previews

I built the Sidekiq job that generates previews for links sent in conversations, with SSRF protection, Redis caching and an automatic pause for slow sites. Tested at 10,000 links per hour, it ensures that a slow third-party site doesn't delay the system's main queues.

10klinks per hour in testing

Ruby on RailsSidekiqRedis
auto_assignment.rb

NChat

Automatic conversation assignment

I developed automatic conversation assignment in three modes (round-robin, by volume and by workload), with the workload score computed in SQL and the public API updated. This way, work is shared more fairly among agents.

Ruby on RailsPostgreSQL
forward_messages.rb

NChat

Forward messages and send contacts

I implemented forwarding a message to up to five conversations and sending contact cards over WhatsApp, in a REST endpoint with validations integrated with two providers. The result is faster customer service, with no copying and pasting between conversations.

Ruby on RailsREST APIs
php_api.php

Dakota Calçados

PHP API from scratch

At Dakota Calçados, I built a PHP API from scratch and handled maintenance and code optimization, working with several departments across the company. The result was better-performing systems that are easier to scale.

PHP
guesty_oauth.ts

NLearn

OAuth2 integration with Guesty

I integrated NLearn with Guesty, a short-term rental management system, via OAuth2, with token caching and a token issued only once, even under concurrent calls. This opened the product to that system's customers with no risk of exceeding the API quota.

TypeScriptOAuth2PostgreSQL
contract_negotiation.ts

NHub

Contract negotiation with approval

I built the contract negotiation flow, in which a discount is proposed, approved or rejected and then synced with Conexa, the billing system, with new endpoints and RLS. This way, the negotiated price is the same one that gets billed, and the sales team became more efficient.

TypeScriptPostgreSQLRLS
whatsapp_meta_templates.rb

NChat

WhatsApp templates connected to Meta

I built the WhatsApp template builder inside the dashboard, connected to the Meta API, where customers create message templates and submit them for approval. This way, the whole process happens without leaving the system.

Ruby on RailsMeta API
support_chatbots.json

Automations

AI customer service chatbots

I developed AI chatbots using Typebot and n8n, connected to the systems' APIs, that answer customers' common questions instead of a support agent. They cut human support volume by up to 60%.

60%less human support

n8nTypebotJavaScript
ai_that_cites_sources.ts

NLearn

AI that cites its sources

I tuned the AI knowledge extraction to require a quote from the conversation for every claim and a self-review before answering, because the model was filling gaps with steps nobody had mentioned. As a result, the content suggested for the knowledge base became trustworthy.

TypeScriptLLMs
daily_sync.ts

NHub

Reliable daily sync

I fixed the daily data collection, which was wiping data and marking active customers as cancelled: the reader now accepts two formats in Redis, the status is only written once the source confirms it, and the API pagination was rewritten. The dashboard's numbers and alerts became reliable again.

TypeScriptRedis
account_isolation.sql

NLearn

Account isolation

I strengthened isolation between accounts in PostgreSQL, with account checks inside functions, a generated allowlist and RLS on views and partitions. The blocking is proven by automated tests, and one account's data stays invisible to the others.

PostgreSQLRLSSupabase
financial_management.tsx

NHub

Financial management platform

I developed the financial management platform that brings together each customer's contract, properties and usage in one place, presenting the data clearly. As a result, I reduced the finance team's information-gathering work.

ReactDenoPostgreSQL
sales_module.tsx

NHub

Sales module

I built the end-to-end sales module in React, on top of Deno Edge Functions and PostgreSQL: a pricing calculator with volume discounts, upsell analysis and a sales proposal, covered by 148 automated tests. The sales team can now simulate, negotiate and generate the proposal in one place, speeding up sales and upselling.

148automated tests

ReactDenoPostgreSQL
demo_mode.tsx

NLearn

Demo mode

I created the demo mode for Insights, with sample data, variants by account state and a seven-step guided tour, so new accounts don't see an empty screen. As a result, the sales team got a screen to present the product.

ReactTypeScript
campaign_cost.rb

NChat

Estimated campaign cost

I implemented the estimated cost of each campaign before it goes out. As a result, customers know how much they'll spend before sending and avoid surprises on their bill.

Ruby on Rails
booking_details.rb

NChat

Booking next to the conversation

I placed the booking and property details next to the conversation, pulled from the customer's management system. This way, agents reply to guests without switching systems, and customer service gets faster.

Ruby on Rails
profile_discounts.tsx

NHub

Discount management

I created discount management on the customer profile. As a result, the sales team tracks and adjusts each customer's terms in one place.

ReactTypeScript
pdf_report.ts

NLearn

PDF report

I developed the PDF report by property and period, with customer service insights. This way, customers take their operation's numbers out of the system and share them with whoever needs them.

TypeScript
process_automations.js

Automations

Process automations

I developed automations for internal processes and guest messaging with JavaScript and n8n, connecting APIs and databases, so that manual processes started running on their own, quickly and effectively.

JavaScriptn8nREST APIs
pull_request_qa.md

New Byte

Pull request QA

I run QA on pull requests from other developers, in Ruby and JavaScript, reproducing every case in the test script before approving. This way, defects are caught before they reach the customer.

RubyJavaScript
production_fixes.md

NChat

Production fixes

I investigated and fixed production defects: the webhook that switched off the customer's bot by mistake, four real-time notification failures and permission-restricted buttons that now explain why, as well as unblocking the deploy at the security check. This way, the system stayed stable for customers.

Ruby on Rails
audit.md

NHub

Critical audit issues

I resolved ten critical issues flagged by an audit of the system. As a result, the product became more secure and reliable for customers.

TypeScriptPostgreSQL
pre_deploy_quality.yml

NLearn and NHub

Quality before deployment

I added visual regression and accessibility tests to CI for NLearn and a quality stage before every deploy for NHub. This way, problems are caught before they reach production.

PlaywrightCI
shopping_app.py

Personal project

Shopping App

I developed an online store with session-based login, a per-user cart and an admin panel for adding products, with the back-end in FastAPI and a SQLite database accessed through SQLAlchemy. It was my project to practice back-end development in Python, from routes to access rules.

PythonFastAPISQLite
See on GitHub ↗
scheduler.java

College

Process scheduler

Working with a partner, I built a process scheduling simulator in Java, with two ready queues with different priorities and time quanta and queues for disk, tape and printer. The Operating Systems assignment shows, tick by tick, how the system decides who uses the processor.

Java
See on GitHub ↗
experience02

52 PRs in production in 2026

and shipping more

Experience and education

Three jobs and a degree in progress.

work
Jun 2025 → present

New Byte

Full-time · remote

  • Junior Software EngineerAug 2026 → present
  • Systems DeveloperSep 2025 → Aug 2026
  • DEV Intern · Implementation and SupportJun 2025 → Sep 2025

Back-end for three products in production.

Sep 2024 → Mar 2025

Dakota Calçados

Software Engineering Intern

A PHP API built from scratch.

2021 → 2022

Sicredi Pioneira

Young Apprentice

education
Feb 2025 → present

Feevale University

Bachelor's in Computer Science

Aug 2023 → Feb 2024

North Broward Prep School

Exchange program in Florida, fluent English

journey03

GitHub activity

356

contributions in the last 12 months

0Dec
0Jan
44Feb
70Mar
52Apr
4May
1Jun
20Jul
26Aug
43Sep
39Oct
56Nov
github04

Let's talk

Interested in my work, a project or an idea? Send me a message! :D

contact05

© 2026 Rafael Lehnen MedeirosRS