AI2 min read
Multi-tenant RAG: isolating client data in AI agents
The problem When you build an AI agent for one client, RAG is straightforward: embed their docs, store them, query…
2026-08-30
// how it's built
Engineered to hold up.
Not buzzwords — the actual architecture decisions behind every build.
01
Scalability & performance
Fast apps that hold up under real traffic, not just a demo.
Proof: Built on Next.js + Cloudflare Workers — edge-deployed, no cold-start server to scale.
02
Security
Data protection and secure-by-default practices, not bolted on after launch.
Proof: Per-tenant data isolation, auth, and roles from day one — see Melow.
03
Automation & DevOps
Continuous deployment with zero downtime — this site is the proof.
Proof: This site ships itself: cron-driven content, one-click approvals, zero-downtime deploys to Cloudflare Workers.
// selected work
Built and shipped.
A few of the integrations, apps, and systems I've shipped — each one solving a real problem and shipped.
Notes
Things I hit while building
Real constraints, tradeoffs and things that broke — the short version of the work the case studies don't tell.
AI2 min read
Getting AI images to write text correctly
The problem Every AI image model I've used mangles text. Quotes, numbers, names — they come out as gibberish far more…
2026-08-30
AI2 min read
Structuring clinical notes in production with an AI copilot
The problem A dentist I work with spends half the consultation typing notes into a patient record. The other half,…
2026-08-30
The toolbox
One builder. Six systems.
Every layer of the stack in one set of hands — AI to motion, connected and shipped.
Explore the services →Focus or hover a system to inspect it — or jump straight to the services.
Next.js 16FlutterCloudflare WorkersD1DeepSeekPostgreSQLBlender
// what I build
One developer. The whole stack.
No handoffs, no account managers. You talk to the person writing the code.
Launch your MVP
A working product in weeks, not quarters — so you validate before you over-build.
Best for
Founders validating an idea with real users, and teams that need a first version fast.
What you get
- ›End-to-end build: product, backend, and the boring-but-critical infrastructure
- ›Weekly working builds you can click — not status decks
- ›Launch-ready deployment with monitoring and a clean handover
Turnaround: First release in 2–6 weeksProof: MoneyGuard · Melow
Automate the manual work
Turn the repetitive weekly grind into pipelines that run themselves.
Best for
Ops leaders whose teams retype, reformat, and reconcile the same data every week.
What you get
- ›WhatsApp/Telegram bots that answer from your own documents
- ›Social publishing and data reconciliation that run on a schedule
- ›Cost controls and fallbacks, so an outage degrades instead of breaks
Turnaround: First automation live in 1–2 weeksProof: Social Command Center
Put AI to work on your data
Assistants and copilots grounded in your documents — that stay quiet instead of hallucinating.
Best for
Businesses with documents, tickets, or records they want turned into answers and automated work.
What you get
- ›RAG grounded in your own data, not a generic script
- ›Human handoff that actually works
- ›Provider-agnostic, so you're never locked to one vendor
Turnaround: 2–6 weeks depending on depthProof: Melow — clinical copilot
// about
I ship production software, solo — fast, without the agency overhead.
Based in Santiago, Dominican Republic. I've shipped 8 projects this year alone — AI integrations, multi-tenant SaaS, mobile apps, and automations. Bilingual EN/ES, and I build for the Dominican market as fluently as the US one.
If you need a smart layer between your data and your users, that's exactly what I build.
More about Pedro →
8
projects built
1
developer — you talk to me
EN/ES
fully bilingual delivery
24h
typical reply time