- Built the mobile app from scratch in React Native and Expo, and have been behind every release and feature of it.
- Shipped an OpenAI support-ticket classifier that routes by reason in production, scored daily against human experts. Over 90% accuracy.
- Took a WhatsApp integration on the Meta Business API from investigation and ADR through production rollout. Account-activation tickets down 63% in 3 weeks.
- Built voice entry for the health and habits section of the medical record, so a member can speak those answers instead of typing them on a phone, and photo capture that reads a medicine box's name, presentation and concentration.
Lucas Moreira e Silva Alves
João Pessoa, Paraíba, Brazil
Languages: Portuguese, English, Spanish
+55 (83) 99106-7863•lmsa.moreira@gmail.com
lucasmsa.com•github.com/lucasmsa•linkedin.com/in/lucasmsa
Software engineer with six years building products end to end, most recently a senior care app taken from its first commit to the AI features running in it.
Experience
- Replaced Create React App, Webpack and Babel with Vite and SWC, and dropped 80+ stale dependencies. Boot time down 80%, production builds down 60%.
- Migrated the front-end to TypeScript and untangled deeply coupled modules, deleting 5,700+ lines of JavaScript and removing 1,500+ circular dependencies.
- Wrote 2,000+ RSpec tests across features in a mature Rails codebase, about 10% of all test additions.
- Shipped critical production hotfixes in under an hour with Datadog and Airbrake, and acted as the team's front-end reference in reviews and architecture.
- Built screens and features on the Mirae platform in Next.js, backed by Strapi as a headless CMS.
- Shipped features for the Livelo Marketplace mobile app, with Storybook and Atomic Design as the component system.
- Delivered a port and vessel management app with offline-persisted sessions via redux-persist, i18n, and Firebase dynamic links and cloud messaging.
- Built a ZeroMQ microservice backend with NLP.js for intent extraction and Danfo.js for structured data.
- Automated Google Calendar scheduling with Puppeteer.
- Cleaned and augmented large datasets with Python and regex for deep-learning training on a Portuguese to Libras translator.
- Contributed to Vlibras-Console for model training and testing, with results published in SEMISH 2021.
Projects
Which 2026 World Cup path carries the worst logistics, group stage through the final, modelling travel, circadian shift, altitude and heat against a base camp. React Three Fiber, Python, Go
Interval training builder for workouts you set round by round, installable and usable offline. React 19, Vite, Zustand, Playwright
Dark VS Code theme, 24,000+ installs on the marketplace.
Drop a screenshot of your editor and it names the theme, in the browser: a CNN embedder exported to ONNX and CIE Lab colour distance over 32k crawled themes. TypeScript, PyTorch, ONNX Runtime Web, AWS S3, React 19
Education
Core Competencies
Frontend: TypeScript, React, React Native, Expo, Next.js, Tailwind, Three.js, Vite
Backend: Laravel, Ruby on Rails, Node.js, Python, GraphQL, REST
Data & Infra: PostgreSQL, Docker, AWS, Vercel, PostHog, Sentry, Datadog
AI & Agents: LLM APIs, MCP, PyTorch, ONNX Runtime
Interests: Design Engineering, AI Engineering, Product Engineering, Agentic Workflows