Software engineer from São Paulo, Brazil. Curious from birth, learning how the world works from the inside and from the outside — and expressing it with technology: agent infrastructure, open source, and streaming systems for hundreds of millions of people.
- Contributing to OrchestratorInc/agent-orchestrator: native ZCode chat integration and Hermes Agent harness with ACP chat, both in review
- Building on Hermes Agent: skills, plugins, loop and harness engineering
- Rebuilding my home on GitHub — The Star Map (every project delivered, mapped onto the real night sky) plus a new blog at
wesleymatosdev.github.io/blog
54 merged PRs across 16 repositories. A selection:
| Project | What I contributed |
|---|---|
| rust-lang/rustfmt | semver support for required_version per the Rust Book spec (#6066, #7048) |
| mediar-ai/screenpipe | early contributor — 22 merged PRs: CI/CD pipelines, Deno & Bun SDK test automation, multi-monitor support with real monitor names, H.265 capture on macOS, Sentry error tracking, automated release notes |
| calcom/cal.diy | QA and test automation for payments, calendar and webhooks |
| triggerdotdev/trigger.dev | ESLint plugin enforcing task-key uniqueness; CLI test suite |
| nashaofu/xcap | real-time tracking of open apps and connected monitors, via Apple APIs |
| qdrant/qdrant | read-only API keys |
| zio/zio-json | 256-bit integer decoding |
| permitio/docs | Docusaurus v3.7 docs upgrade |
Plus one-off merged fixes in Handy, ts-jest, firstmate, Cap-go, inb, zcode-cli, ai-memory and marinara.
- Warner Bros. Discovery (via Globant) · Senior Software Engineer → Staff Cloud/SRE Engineer — streaming infrastructure behind
hbomax.com,cnn.comanddiscoveryplus.com: hundreds of millions of users (95M on HBO alone), multi-region multi-cloud CDNs toward a 99.99999% availability target, GraphQL at 15–25M requests/day, Kafka/SQS content pipelines, a legacy-CMS replacement saving $5M/yr, and 100+ CVEs remediated across 50+ repositories - Influencer · Staff Software Engineer — creator-discovery platform fusing AI inference with creator, audience and demographic filters; vector search over AI embeddings; cut third-party API latency from 30s to 900ms (97%)
- Luxury Presence · Senior Software Engineer II — Testcontainers running parallel E2E suites against live PostgreSQL, SQS and Redis; 100+ test cases; an SDK replacing 10k+ lines of duplicated logic across six NestJS microservices
- Before that: fintech and e-commerce engineering in São Paulo (Provi, FCamara/Dasa, Marketdata/Cielo, RedFox) and teaching 200+ students HTML, CSS, JS, React and Node with Meta's Hack Station
- awesome-acp — curated resources for the Agent Client Protocol
- hermes-orchestrator — merge-friendly agent-orchestrator fork with first-class Hermes support
- hermes-classifier-mode — Claude Code's auto permission mode reviewed by a local LLM (Ollama, on-device)
- ai-usage-optimizer — routes AI provider and model usage by measured headroom
- lolcompanion — lightweight League of Legends companion, Wails v3 + Go over LCU and Live Client APIs, no injection
- awesome-brazil — civic tech and open data in Brazil 🇧🇷
- bela-portfolio — a client's portfolio
- Languages JavaScript · TypeScript · Rust · Go · Python · Node.js · Deno · Bun
- Web React · Next.js · Vue · NestJS · GraphQL · Fastify
- Infra & data AWS · GCP · Kubernetes · Terraform · Kafka · PostgreSQL · Redis · Elasticsearch · Fastly · Akamai
- Agent stack Ollama · MCP · Qdrant · embeddings & vector search · agent skills and plugins · spec-driven development · harness/loop engineering
✉️ [email protected] · 💼 linkedin.com/in/wesleymatosdev · 🌌 The Star Map