10+ years shipping code across very different stacks — APIs, real-time systems, AI agents, embedded firmware, OSINT pipelines, custom DSLs, game-server reverse engineering, whatever pulls me in. Started in .NET/C#, fell into React/Node.js, now I live in TypeScript and run a self-hosted intelligence grid for agentic work.
Self-taught — learned by reading docs, breaking things, and obsessing over how stuff works under the hood. Most of what I build ships under @tecnomancy, my personal R&D org for agentic AI, OSINT and tooling.
|
Languages
|
|
|
Frontend
|
Backend
|
|
Databases & ORMs
|
Testing & Validation
|
|
Cloud (AWS)
|
Self-hosted Infra
|
|
DevTools & Editor
|
|
|
AI & Agents
|
|
- MAGO Intelligence Grid — multi-agent orchestration, custom MCP servers, board-driven SDLC, agent dispatch farms with WIP guards and per-agent telemetry
- Agentic AI — voice-first agents (Gemini Live + Claude Code), browser automation with magic overlays, MCP server design
- OSINT & Security — recursive recon chains, financial intelligence (FININT), leak hunting, secret scanning, wallet tracking, threat intel correlation
- Functional Programming — fp-ts, Effect, algebraic data types, type-level inference (everything inferred, zero deps)
- Self-hosted Infra — Bun-first stateless microservices, MAGO Cloud (own serverless-style runtime), MinIO/S3, OpenTofu
I think a lot about how AI agents do work — not just what they output. Building, breaking and studying the layer between models and reality.
- Agent Harness Engineering — designing the runtime around AI agents: tool surfaces, context windows, memory layers, permissions, hooks, subagent/forking strategies, output economics, MCP federation. Close reading of harness internals (Claude Code, OpenCode) and replicating patterns in my own runtimes.
- Spec-Driven Development (SDD) — specs as the single source of truth (GitHub Spec Kit + custom workflows). Agents generate, verify and refactor code from specs; humans review the spec, not the diff.
- Test-Driven Development (TDD) — tests-first / contract-first. Critical when AI agents drive implementation: a failing test is the only honest signal in a sea of plausible-looking code.
- Custom MCP Servers — protocol bridges so agents manipulate boards, browsers, designs and infra without breaking tool-call context. Designed for narrow, composable surfaces over monolithic god-tools.
- Symphony SDLC — board-driven cycle for agent-orchestrated work: Backlog → Todo → In Progress → Eye QA → PR → Done, with WIP guards, agent telemetry and a hard "single source of truth" board.
- SIGIL — Structured Intelligence Grid Implementation Lifecycle for multi-agent feature delivery, from sigil intake through deploy.
- Currently studying — context engineering, prompt-caching economics, multi-agent voice protocols, deterministic-vs-stochastic orchestration boundaries, formal specs as compilation source over code.
MAGO ecosystem & AI agents
| Project | Description |
|---|---|
| alchemy | Functional programming primitives for TypeScript — every type inferred, zero dependencies |
| magic | MTG AI Judge — deterministic rules engine + semantic RAG over Comp Rules, IPG, MTR & JAR |
Apps & systems
| Project | Description |
|---|---|
| socket-whatsapp | WhatsApp Web socket integration — real-time messaging automation in Node.js |
| take-home | Fullstack monorepo — Next.js 15 + NestJS + Prisma + Redis + React Query + Docker Compose |
Hardware & systems-level
| Project | Description |
|---|---|
| RFID | ESP8266 RFID gateway — MFRC522 reader with LCD/LED feedback and SPIFFS-backed offline queue |
| rust-performance-test | Rinha de Backend 2023 — Rust submission |
| Domain | What I work on |
|---|---|
| Agent Engineering | Harness design, MCP server federation, context engineering, subagent orchestration, voice-first interfaces |
| OSINT | Recursive recon chains, financial intelligence, sanctions/PEPs/offshore lookup, leak hunting, wallet tracking, secret scanning, threat intel correlation |
| Real-time | WebSocket protocols, voice (bidirectional audio), live presence, reverse-engineered web protocols |
| Functional Core | TS type-level inference, ADTs, Effect, fp-ts, algebraic primitives |
| Self-hosted | Single-node architecture, custom serverless runtime, MinIO/S3, OpenTofu, vertical scaling |
| Embedded | ESP8266/Arduino firmware, SPIFFS, RFID, hardware feedback (LCD/LED/buzzer) |
| Game Servers | PristonTale server reverse engineering, C++ server tooling, MTG rules engine |
| DSL & Tooling | Custom languages, GLSL shaders, Neovim config (J.A.R.V.I.S), Ghostty HUD experiments |
| Period | Focus |
|---|---|
| 2016-2018 | C#, ASP.NET MVC, Entity Framework |
| 2019-2020 | React, Node.js, WebSocket, APIs |
| 2021-2023 | TypeScript, Microservices, Monorepos |
| 2024-present | Agent Harness Engineering, AI Agents, OSINT, Self-hosted Infra |
120+ repositories · Self-taught since 2016