One engineer, complete systems: a multi-tenant SaaS platform, an open protocol and a sensing stack β all built solo, front to back.
AI Engineer Β· Applied AI Β· Agentic Systems Β· Python Automation
Lucky Cat, a multi-tenant restaurant platform, built solo front to back. Alongside it, two open-source systems: SCPE, a signed provenance protocol, and Wavr, explainable multi-modal sensing.
65 tables 120 row-level security policies 26 edge functions 144 migrations built solo in 6 weeks
Usually replies the same day Β· remote, Ireland/EU
A multi-tenant SaaS platform carrying three products, a RAG sales copilot, a multi-agent orchestration case study, an open provenance protocol and a privacy-first sensing stack. Built solo, front to back.
One multi-tenant codebase on Supabase RLS, Cloudflare Pages and a Stripe Connect billing integration, carrying three separate products, each with its own users, its own surface and its own tenant isolation. Built solo, front to back.
Built on a RAG pipeline (Gemini + Supabase pgvector): when a field rep hits an objection mid-pitch, semantic vector search surfaces the best context-aware counter in under a second. White-label: rebrandable for any industry in under a day, shown in the case study under a fictional energy-sector brand.
A multi-agent Claude pipeline (research β draft β edit β export) that produced an 84-page book (PDF + EPUB, ~21,000 words). Shown here as a write-up: the pipeline source is private (the book ships commercially), with a free public sample linked below.
Fuses six sensing modalities (network scan, BLE proximity, camera pose via YOLO, mmWave radar and more) into one explainable per-room state (confidence = strength: trust weight Γ source confidence Γ freshness decay), on a top-down radar over a floor plan you draw in-app. A read-only MCP server exposes that state so your own agents can ask "who's home" as structured context. Privacy-first by construction: loopback-only, zero cloud, camera frames never stored, positions never written to disk. Every sensor is mock-tested, so the suite runs with no hardware attached. AGPL, built solo.
When a pull request arrives from someone you don't know (a person, or increasingly an AI agent), trust rests on a username and reading the diff by eye. SCPE adds a signed envelope that proves who produced a contribution and that nothing was tampered with, verified offline with no protocol server and no new accounts, using signing keys the contributor's git host already publishes. I wrote the spec and three independent verifiers in Python, Go and Rust that must reach the same verdict across 18 normative test vectors. Ships as a GitHub Action that seals or gates pull requests. Apache-2.0, on PyPI (v0.2.3).
The stack behind the projects above.
Brazilian builder based in Limerick, Ireland. I learned Python, Supabase, Cloudflare and Claude orchestration by shipping Lucky Cat β a full multi-tenant restaurant SaaS I built solo, front to back, in under two months.
The Graphic Design degree shows in the UI: spacing, contrast and layout are deliberate. I work fast, document clearly, and don't leave clients dependent on me for basic maintenance.
AI automation, LLM integration, Python scripting. Short builds or long engagements, both fine. I usually reply the same day.
Email me