PROFILE
I’m an End-to-End Engineer at CoGrader, building software since 2013. I own initiatives from understanding customer needs and helping define the solution through implementation, release and follow-up. My work combines product decisions, technical leadership and hands-on AI engineering.
At CoGrader, I built a browser extension that reduced integration friction across six learning platforms. It helped attract customers and generate revenue, supporting the growth of our engineering team. I also help build Cortex, with personalized AI feedback, Go/Node.js document processors and reliable background jobs.
CORE COMPETENCIES
System Design & leadershipCustomer problem framing, product and engineering trade-offs, end-to-end delivery, technical leadership and mentoring. API design, asynchronous workflows, queues, idempotency and failure recovery. Languages: English and Brazilian Portuguese.
AI engineeringLLM orchestration, prompt engineering, evals, RAG, classification; Gemini / Vertex AI, OpenAI, Anthropic.
Backend & dataNode.js, TypeScript, Go, Python, Java, NestJS, Express; Redis / BullMQ, MongoDB, PostgreSQL, Firebase / Firestore, BigQuery / Dataform.
Frontend & integrationsReact, Next.js, Angular, Vite, Astro, Tailwind CSS; Plasmo / Chrome MV3, REST, SOAP, WebSockets and payment APIs.
Cloud & deliveryGCP, AWS, Azure, Docker, Kubernetes, Terraform, GitHub Actions, CI/CD, automated testing and idempotent processing.
Observability & AI-assisted developmentSentry, PostHog, Axiom; OpenAI Codex, Claude Code, custom skills and MCP integrations, including PostHog MCP.
PROFESSIONAL EXPERIENCE
- File processing: Built file processors with Go and Node.js for multiple formats, including PDFs, CSVs and images, with support for scanned pages and handwritten student work. Redis-backed queues and idempotent processing let jobs retry without repeating their effects.
- AI orchestration: Orchestrated focused agents and used classification for decisions that do not need an LLM. Gemini runs through Google Cloud Vertex AI, supported by credits received from Google.
- Customer and business impact: The extension reduced integration friction, helped attract customers and generate revenue, and supported the growth of the engineering team. Cortex adapts grading and feedback to each teacher’s style.
- Browser extension: I owned the browser extension initiative from adoption barriers through delivery. Built it from scratch with React, TypeScript and Plasmo, integrating Canvas, Blackboard, Schoology, Brightspace, Microsoft Teams for Education and Google Classroom. Bringing CoGrader into teachers’ existing workflows made the product easier to adopt and helped attract customers and generate revenue.
- Product experience: Contributed to notifications, backend services and the React application. Follow releases through production use, using customer feedback and observability to improve reliability and the teacher experience. Our data stack at CoGrader includes BigQuery.
Backend & file processing: Go, Node.js, TypeScript, Express, Redis, BullMQ, Idempotent jobs, Firestore/Firebase Admin, Python, Zod
Frontend: React, Vite, Astro, Mantine, in-house design system, Tailwind, Plasmo, TypeScript
AI & integrations: Gemini / Vertex AI, Small-agent orchestration, Classification, OpenAI, Anthropic/Claude, Google Classroom, Canvas, Blackboard, Schoology, Brightspace, Microsoft Teams for Education, EdLink, Stripe
Infrastructure & observability: Google Cloud, Azure, AWS, Terraform, Docker, Turborepo, pnpm, GitHub Actions, BigQuery, Dataform, Axiom, PostHog
- Built backend APIs with NestJS and MongoDB for an iGaming CRM, incorporating AI capabilities.
- Built GOATVPN infrastructure with WireGuard. The platform included a React/Next.js frontend, MongoDB Atlas, Redis, payment integrations and automated server provisioning.
- Developed a platform that turned TVs in small and medium-sized businesses into advertising displays, using data and automation to support ad delivery.
Core stack: NestJS, MongoDB, WireGuard, React, Next.js, Redis, AWS, Google Cloud, Vercel, New Relic
- Connected smart displays to Vertex AI, Firebase, Cloud Run, Cloud Storage and MongoDB Atlas.
- Combined dynamic advertising with audience analytics using YOLO, VisionAI and LLMs.
- Reached the first commercial installation at Centro Profissional.
- Received third-place recognition at Adapta 2025. Awarded R$40,000 at the Hackathon and R$100,000 at the Summit, both in cash.
- Installed 100 Android tablets in elevators around São Paulo for the advertising rollout.
Core stack: Android, Google Cloud, Vertex AI, Firebase, Cloud Run, Cloud Storage, MongoDB Atlas, YOLO, VisionAI, LLMs
- Developed and optimized applications with React and Ant Design, backed by NestJS, MongoDB and RabbitMQ. Used test-driven development and Docker to support delivery.
- Led Individual Development Plans to help engineers identify growth goals and develop their skills alongside project work.
- Presented a ChatGPT integration with WhatsApp using Meta’s official API and NestJS.
Core stack: React, Ant Design, NestJS, MongoDB, RabbitMQ, Docker, Git, TDD
- Developed Java backend services with PostgreSQL for bank integrations.
- Used AWS, Google Cloud and Azure infrastructure to support IP rotation for web scraping.
Core stack: Java, PostgreSQL, REST APIs, Web scraping, AWS, Google Cloud, Azure, Docker, Git
- Used Amazon S3 to host Angular applications and AWS CodePipeline for deployment.
- Worked with Java, Docker and Jenkins in the backend and delivery toolchain.
Core stack: Angular, Java, Amazon S3, AWS CodePipeline, Docker, Jenkins
- Developed the web application with Angular and Bootstrap, backed by NestJS and PostgreSQL. Used Amazon S3 for storage and Heroku for hosting.
- Connected the content marketplace to Binance Smart Chain to support NFT transactions.
Core stack: Angular, Bootstrap, NestJS, PostgreSQL, Amazon S3, Heroku, Binance Smart Chain
- Helped build the platform with Angular, NestJS and PostgreSQL, using Docker for containerized development.
- Worked with Git for version control and Scrum with Trello.
Core stack: Angular, NestJS, PostgreSQL, Docker, Git, Scrum
- Built web interfaces with Angular and Angular Material, and worked with Java and Node.js on the backend.
- Worked with Git, Jenkins and Scrum with Jira in the delivery workflow.
Core stack: Angular, Angular Material, Java, Node.js, Jenkins, Git
- Worked with REST APIs, Nginx and Git as part of the application and development stack.
Core stack: PHP, Symfony, REST APIs, Nginx, Git
An independent fantasy game and its surrounding product: gameplay, player accounts, cloud saves, purchases, a web portal, infrastructure, art and audio. I work across the full journey, from building the experience to preparing releases.
I build Natiely as an independent fantasy game with Unity 6 and C#. The project brings together procedural worlds, character movement, combat, enemies, progression and saves. Unity URP, Shader Graph, VFX Graph, Cinemachine and the Input System support the visuals, effects, camera and controls.
I work on downloadable world content with Unity Addressables, download progress, caching and recovery, alongside asset optimization and mobile builds. The goal is to keep large environments manageable while preserving the player's progress when a download or connection fails.
The companion website and player portal use React, TypeScript, Vite, Tailwind CSS and shadcn/ui, with TanStack Router, TanStack Query, Zustand and Zod. My work spans the public experience, account flows and store interfaces. The repository also includes an earlier Tauri/Three.js game prototype.
I develop Go APIs with Echo, MongoDB and Redis, integrating Firebase authentication and Google/Apple sign-in. Cloud-save synchronization uses revision checks and idempotent writes to handle retries and conflicting updates across devices. Local gameplay and remote account services have separate responsibilities.
I integrate RevenueCat for native purchases and Stripe for web checkout, with backend purchase verification, webhook processing, Gold balances and coupon flows. Firebase Realtime Database supports wallet updates. Store-specific payment rules stay separate from the game's own economy and interface.
I use Sentry for technical errors and diagnostics, PostHog for web product analytics, and OneSignal for player notifications. I investigate user behavior through PostHog MCP and connect that evidence with technical diagnostics to prioritize fixes and improve the experience.
My work includes Docker, Google Cloud, Cloud Storage, AWS infrastructure and GitHub Actions deployment workflows, plus Fastlane tooling for Android and Apple publishing. I handle build configuration, asset delivery, signing workflows and store materials as part of bringing the product to players.
I combine Blender and Blender MCP with Meshy for 3D asset work, generate sound effects through the ElevenLabs API, and create music with Suno. OpenAI Codex, Claude Code, custom skills and MCP integrations support development, investigation and iteration across the project.
HACKATHON PROJECT / CLASSGAME
Built a 3D learning experience for the All Things Agentic Hackathon, combining adaptive AI tutoring and a teacher dashboard with Next.js, Three.js, Go and Gemini on Vertex AI.
Public repository: github.com/velrino/classgame
CREATIVE PRACTICE & DEVELOPER WORKFLOW
Games are my personal hobby and the starting point of my self-taught approach to software. Beyond Unity, I build interactive WebGL experiences with Three.js, including my playable portfolio. My creative tools include Blender and Blender MCP, Meshy for 3D assets, ElevenLabs API for sound effects and Suno for music.
I use OpenAI Codex and Claude Code with custom skills and MCP integrations for investigation, implementation and review. PostHog MCP helps me understand user behavior; Sentry and Axiom provide technical diagnostics to guide fixes.
PRESENTATIONS & KNOWLEDGE SHARING
Organized meetups and workshops to help beginners learn software development. Presented ChatGPT in WhatsApp, covering integration through the official WhatsApp API with NestJS.