Brayan MartinsBM

Full-stack developer · Santa Maria, Brazil · @BragaNux

Photo of Brayan Martins

I build systems end to end, from the database to the mobile app. I've been a full-stack developer since 2024. I co-founded an incubated startup and I'm in the final stretch of my Information Systems degree.

Open to remote opportunities, in Brazil and abroad

See projects Résumé PDF ↓ LinkedIn ↗ GitHub ↗

Career timeline

Each dot is a repository I worked on, placed at its last update and colored by its main language. The bands show where I was at each point. Hover over or tap a dot to see the project.

What's in the code

  • C# 18%
  • TypeScript 17%
  • Dart 11%
  • Rust 10%
  • Python 9%
  • Go 8%
  • JavaScript 8%
  • SQL 6%
  • HTML/CSS 5%
  • Java 5%
  • Infra, Assembly and other 3%

Based on 1,380 code files across the repositories, including team projects.

See all 49 repositories

Coursework exercises are collected in estudos-faculdade.

UpdatedRepositoryWhat it isLanguageCode

Featured projects

Five projects that show how I think through a whole system: the problem, the architecture and what was delivered.

2025 – 2026Capstone · authorDue Nov 2026
Capstone project20 of 20 API tests

Panela

A group expense-splitting app that helps people plan before they spend, a step Splitwise and Tricount leave out. It has budgets by spending category, cost splitting, settlement through Pix, the Brazilian instant payment system, and an AI that sums up where the group stands.

Panela welcome screen with the tagline Divida a conta, não a amizade, or split the bill, not the friendshipPanela welcome screen with the tagline Divida a conta, não a amizade, or split the bill, not the friendship
Welcome
Home screen with the balance between friends, a pending Pix payment and open groupsHome screen with the balance between friends, a pending Pix payment and open groups
Balances and debts
Planning screen with an event, the total budget and reserves by spending categoryPlanning screen with an event, the total budget and reserves by spending category
Planning
AI-generated group summary explaining who owes how muchAI-generated group summary explaining who owes how much
AI summary
  • Four ways to split: equally, by percentage, by fixed amount or payer only. Settlement suggests the fewest possible transfers.
  • Planning with budgets and reserves by spending category, with push alerts at 80% of the budget and when it goes past 100%.
  • AI to suggest budget items, read receipts and summarize the group in plain language.
  • Automated functional API tests: 20 of 20 cases and 90 checks, including the permission rules.

How it works

Flutter app→ASP.NET Core API→PostgreSQL
Push via FirebaseAIPlace search (OpenStreetMap)Docker + Caddy

Stack

Flutter · DartASP.NET Core 8EF CorePostgreSQLJWTFirebase MessagingDocker · Caddy
  • Dart 54%
  • C# 25%
  • TypeScript 17%, from the earlier React Native version
Panela_Rateio ↗
2026Product · co-founderIn development
Nexa Solution team · incubated at InovAMFPrivate repository

3Summa

A product for accounting firms that I'm building with my two partners on the Nexa Solution team. It's built around client service over WhatsApp, with a shared inbox for the team and real-time cost tracking in Brazilian reais. Within the team, my focus is the front end.

  • Incoming messages arrive by webhook with HMAC-SHA256 signature validation and idempotent writes, so the same event is never processed twice.
  • Cost dashboard in reais, with configurable rates and an automatically updated dollar exchange rate.
  • Agent accounts with Argon2id, revocable sessions, access roles and email-based 2FA.
  • Per-client data isolation in PostgreSQL is on the roadmap, on the way to becoming SaaS.

How it works

Client on WhatsApp→Webhook with HMAC→Rust API→PostgreSQL→React inbox

Stack

React 19 · TypeScriptRust · Axum · SQLxPostgreSQL 17WhatsApp integrationOpenAPI · Swagger
  • Rust 47%
  • TypeScript 37%
  • SQL 15%
2026Code Race AMF · Nexa Solution teamCompetition
Best usability awardPrivate repository

Agribusiness SaaS

A management system for the agribusiness chain in Rio Grande do Sul, from farmer to cooperative. What sets it apart is an AI agent on WhatsApp that looks up and records data by message.

  • LangChain agent using Llama 3.3 70B (Groq) and Claude, with tools for inventory, batches, weather and semantic search.
  • Semantic search with pgvector and embeddings, in the same PostgreSQL as the business data.
  • Batch inventory with FEFO and alerts, QR code traceability, and compliance with LGPD, the Brazilian data protection law, through versioned consents.
  • Incremental back-end migration from Django to Rust, keeping the same HTTP contract.

How it works

Farmer on WhatsApp→Node.js service→AI agent (LangChain)→Inventory, batches, weather, search→PostgreSQL + pgvector

Stack

React · TypeScriptDjango 5 → Rust · AxumFastAPI · LangChainPostgreSQL · pgvectorNode.jsNginx · Docker
  • Rust 37%
  • TypeScript 26%
  • Python 22%
  • SQL 12%
2026Team projectPO and front end
Observability from day one

Up Espaço

A platform that connects the staff of an early childhood development center with families: attendance, an activity feed with photos, a daily timeline for each child, developmental milestones, a calendar and announcements.

  • REST API in Go with Clean Architecture and the Repository and Decorator patterns, with no dependency injection framework.
  • Each guardian sees only their own children, and the staff sees every student.
  • Prometheus and Grafana from the first version, with dashboards for requests, errors, CPU, memory and p95 latency.
  • All 10 planned requirements, 5 functional and 5 non-functional, were delivered and demonstrated.
  • I led product and front end: backlog, team organization, budget and benchmarking.

How it works

React SPA→Go API→PostgreSQL
Prometheus→Grafana

Stack

Go · gorilla/muxReact 18 · Vite · TailwindPostgreSQL 15JWT · bcryptPrometheus · GrafanaDocker Compose
  • Go 66%
  • SQL 21%
  • JavaScript 9%
up-espaco ↗Product documentation ↗
2026AcademicTwo implementations
RAGDeployed with Docker and Nginx

BookGuess

A literary Wordle. Every day there's a new excerpt from a classic book; players try to guess the book or the author and can ask for an AI-generated hint, which costs half their points.

  • RAG hints: the API finds similar excerpts in pgvector and passes that context to Claude.
  • Answers are checked on the server with SSR, so the title never reaches the browser too early.
  • Real-time leaderboard with Redis Sorted Sets and a cached daily challenge.
  • Public domain works only. One version in .NET with Clean Architecture and another in Next.js + FastAPI, deployed with Docker, Nginx and SSL.

How it works

Next.js with SSR→FastAPI→pgvector→Claude
Redis for leaderboard and cache

Stack

Next.js 14 · shadcn/uiFastAPI.NET · BlazorPostgreSQL · pgvectorRedisClaude APIDocker · Nginx
  • C# 53%
  • Python 18%
  • TypeScript 12%
  • HTML/CSS 11%
bookguess-dotnet ↗bookguess-nextjs ↗

More projects

Computer vision, machine learning, microservices, mobile, systems analysis and a game, some with demo videos. The full list is in all repositories.

2026 · Computer visionRaspberry Pi

EcoScan

Sorts trash into 12 categories through the camera, in real time, built to run on a Raspberry Pi mounted above the bin. Trained on about 35,000 images from two datasets, it reached about 92% validation accuracy.

TensorFlow · MobileNetV2OpenCVTFLite
ecoscan ↗
2026 · Machine learningTeam of three

NYC Rat Sightings

Predicts whether a rat sighting complaint will be resolved quickly or slowly, combining 311 requests, health inspections and weather. 0.78 AUC on the test set and 84.5% precision on the requests the model ranks as most likely to be slow, with temporal validation to check that it generalizes.

Python · scikit-learnpandasNYC Open Data · NOAA
ratos-nova-york ↗
2025 · Browser extensionChrome

Sentinel AI

From the right-click menu, it analyzes text, links and images and returns the likely authorship, the risk level and the reason. It keeps a history with CSV export and uses its own back end, which switches between Gemini and a local model on Ollama.

JavaScriptNode · ExpressGemini · OllamaDocker
Sentinel-AI ↗
2026 · RAG and automationTelegram

Cooking Assistant

Telegram bot that answers cooking and food safety questions based on PDF manuals, using RAG with a local vector database, orchestration in n8n and alerts on Discord.

Pythonn8nVector databaseTelegram · Discord
assistente-culinaria ↗
2025 · Architecture5 services

Microservices e-commerce

Users, products, orders, payments and notifications as separate services, behind a Kong gateway and talking over Kafka, with metrics in Prometheus and Grafana and load tests in K6.

Node.js · PrismaKafka · KongPostgreSQL · MongoDBK6
ecommerce-microsservicos ↗
2026 · Design patternsJava

BragaFlow Atlas

Java API with MongoDB Atlas for freight and deliveries, organized in layers, with Factory Method for the transport types, Singleton for the caches and a web interface for testing.

JavaMongoDB AtlasFactory Method · Singleton
bragaflow ↗
2026 · Analysis and designTeam project

Restaurant Service System

Analysis and design of a system that digitizes the dining room: ordering from a tablet at the table, automatic routing to the kitchen, real-time table management and payment. Includes requirements, use cases and class, sequence and activity diagrams.

RequirementsUMLUse cases
Academic documentation
2026 · Product proposalReal client

MaestroOS

Digital transformation proposal for an olive oil brand: an online store with Pix, WhatsApp service with an AI sommelier, QR codes showing where the product comes from and a management dashboard with CRM and subscriptions.

Next.js · TypeScriptSupabaseVercel
Client proposal
2025 · MobileExpo

Event Check-in

App for the door staff to check attendees in by scanning QR codes, with haptic feedback and stack navigation.

React Native · ExpoTypeScriptExpo Camera
checkin-evento ↗
2025 · E-commerceFull-stack

UniStore

A store for students with login, a cart and orders saved per user, simulated payment and step-by-step delivery tracking, with a Node.js and PostgreSQL back end.

React + TypeScriptNode · ExpressPostgreSQL
unistore ↗Demo video ↗
2026 · Educational gameGroup project

O Algoritmo

A pixel-art game, in Portuguese, that spans 3,800 years of the history of algorithms, from Babylon to today, and ends with building a machine. It comes with an interactive page on punched cards, with a simulator. Made for a presentation in the Innovation course.

Canvas · JavaScriptWeb AudioVercel
Play ↗Fun facts ↗
2024 · Reinforcement learningPygame

Tic-Tac-Toe AI

A neural network trained with reinforcement learning so it never loses to a human. It started as an AI for truco, a Brazilian card game, and changed course because of the bluffing, a decision documented in the repository.

PythonTensorFlowPygame
ia-jogo-da-velha ↗
2024 · Multimodal AIFlask

Pharmacy Specialist

Web interface that combines audio transcription with Whisper, image description with CLIP, text generation and video processing.

Python · FlaskWhisperCLIP
especialista-farmacia ↗Demo video ↗
2024 · AlgorithmsFlask

E-commerce crawler

Collects products from an online store and sorts them by price, using external sorting in blocks with temporary files for volumes that do not fit in memory. It has a web interface to run the search.

Python · FlaskBeautifulSoupExternal sorting
crawler-ecommerce ↗Demo video ↗

Skills

Tap a technology to see where I used it. The ones marked with a dot are part of my daily work.

Mobile
Web
APIs and back end
Data
AI and machine learning
Infra and quality
Software engineering
Security

Where I used it

C# · .NET

PanelaBookGuessWork at Forza

How I work

Army and college

Discipline and resilience

I served a year in the Brazilian Army and then balanced a full-time job with college.

Two years in logistics

Organization and processes

I handled supply and distribution operations. I bring that sense of flow and deadlines to software projects.

Up Espaço and 3Summa

Leadership and product

I was PO and front-end lead on a team of five, and I'm one of three co-founders of an incubated startup.

Code Race and Panela

Focus on the user

Our team won the best usability award at Code Race 2026. In Panela, the focus on planning before spending came from a gap in existing apps, and the user evaluation will use the SUS scale.

Teaching and research

Communication and teaching

I was a teaching assistant for Algorithms I, published three expanded abstracts at MICAMF and document my projects so anyone can run and understand them.

Day to day

Quality and investigation

I do code reviews, write automated tests and look for the root cause in the data and the version history before touching the code.

Experience and education

  1. 2026 – nowInovAMF · FOIL

    Co-founder · 3Summa

    A product for the accounting sector created by our team, Nexa Solution, and incubated at InovAMF. The idea is to take repetitive work out of the daily routine of accounting firms.

  2. 2024 – nowSanta Maria, Brazil · on-site

    Full-stack developer · Forza Clubes

    I work end to end on a management system used by clubs, across the APIs, the web system and the mobile app. I also built the internal platform the team uses to organize knowledge and documentation.

    C# · .NETASP.NET CoreBlazorFlutter · DartReact · TypeScriptMySQLJWTSVN

    Day to day: code reviews, automated tests and root cause analysis.

  3. 2022 – 2026Antonio Meneghetti Faculdade

    Bachelor of Information Systems

    Capstone project: the Panela app, due November 2026. Teaching assistant for Algorithms I, three expanded abstracts at MICAMF and two Code Race participations, one of them with an award.

  4. 2022 – 2024Santa Maria, Brazil

    Logistics administrator · Modo Logística

    Full-time supply and distribution operations during my first years of college.

  5. 2021 – 2022Santa Maria, Brazil

    Brazilian Army

    A year of military service, where I built discipline and resilience.

  6. 2018 – 2019Santa Maria, Brazil

    Young apprentice · Beltrame Supermercados

    My first job.

Awards, research and courses

Awards, research and teaching

  • 2026Best usability awardCode Race AMF, with the Nexa Solution team
  • 2025Expanded abstract at MICAMFVITALITY, a gamified system for digital well-being
  • 2024Expanded abstract at MICAMFA tactile and audio device for visually impaired people
  • 2024Expanded abstract at MICAMFBlockchain beyond cryptocurrencies
  • 2024Competitor at the 8th Code Race AMF16-hour competition
  • 2023Teaching assistant, Algorithms I Lab40 hours

Courses and languages

  • 2026Introduction to LinuxAlfahelix
  • 2025AluraC programming (two courses), Figma and programming logic with JavaScript
  • 2025Practical AI promptingWriting prompts for artificial intelligence
  • 2024AluraC#, Java, quality assurance and bug management
  • 2022Backend Descomplicando com C#Tech Courses AMF, 16 hours
  • 2022JavaScript, CSS and HTML16-hour course
  • LanguagesNative PortugueseEnglish in progress, aiming for fluency

On GitHub

Each square is a day with contributions over the last twelve months, private repositories included. The snake is redrawn every day by a GitHub Action.

@BragaNuxView profile ↗

I'm open to remote opportunities, in Brazil and abroad, on teams that value code quality, good practices and continuous learning.

Email
devbrayancomercial@gmail.com
WhatsApp
+55 (55) 99205-3272Open in WhatsApp ↗
Résumé
English (PDF) ↓Portuguese (PDF) ↓
Instagram
@b.martlns ↗
LinkedIn ↗ GitHub ↗