codeSHIFT
Independent AI Engineer · Florianópolis, Brazil

I build AI systems that reach production — and stay there.

A decade of production engineering — Salesforce, dLocal, The Sandbox. Today, AI that handles 15,000+ conversations a month at 90%+ accuracy.

Currently AI Engineer at SignalCore AI, building autonomous decision systems for enterprise.

Book a 15-minute call Free · 15 min · no obligation
SignalCore AI·Salesforce·dLocal·The Sandbox
Ramiro González
Who you’re working with

A backend engineer who moved to the AI layer — and ships it.

I’m Ramiro González, a software engineer with a decade building backend systems for companies like Salesforce, dLocal and The Sandbox. For the last few years I’ve worked entirely on AI — taking language models out of the demo and into production, where real people and real load are what break them.

I don’t sell packaged services. I take on a project and go all in: I map how the work happens today, get something running in weeks, and hand you code you own outright. Based in Florianópolis, Brazil, I work across US and European hours.

Florianópolis, BrazilEnglish · Português · EspañolBSc Software Engineering10 years shipping to production
What I do

Five capabilities — each with something shipped behind it.

I’m one engineer, not an agency — so I go deep, not wide. This is the ground I cover when I take a project on.

01

RAG architecture

90%+ answer relevance with citations; 65% fewer hallucinations via semantic chunking and hybrid search.

pgvectorQdrantPinecone
02

Agent orchestration

10+ agents in production, coordinating tools, memory and human handoff.

LangChainLangGraphLlamaIndex
03

LLM systems at volume

15,000+ conversations a month, holding accuracy under real production load.

04

Evaluation & guardrails

Every system ships with LLM-as-judge evals, cost and latency monitoring, and built-in escalation.

05

Backend at scale

120K monthly active users at The Sandbox; cross-border payments infrastructure at dLocal.

How I work

What working together actually looks like.

01

We start with a 15-minute scoping call. Free. If it isn’t a fit, I’ll tell you on that call.

02

I map the process before I write code. Most AI projects fail by automating a broken process faster. First I want to see how the work happens today.

03

Something runs by week three or four, even if it’s narrow. A system you can use badly beats one you can only discuss.

04

I’ll tell you what the AI should not do. Every project has a line where the machine stops and a person starts. Drawing it early is most of the difference between a production system and a demo.

05

You own the code. Repository, infrastructure, prompts — all of it, from day one.

Based in Florianópolis (UTC−3) · hours overlap the US & most of Europe · English · Portuguese · Spanish

Where I’ve applied it

Niches I’ve already shipped into.

If your world is on this list, I’ve built for it before — as a founder’s first AI hire, or as the engineer behind the product.

Aesthetic clinics
AI SDR · flagship
Education & ed-tech
RAG support agent
Finance / fintech
AI learning agent · building
Architecture & AEC
Decision memory · 2 projects
Creative / digital clone
Award-winning · 3 clients
Payments
dLocal
CRM / SaaS
Salesforce
Web3 gaming
The Sandbox
Logistics
XPO Logistics
Case study 01 · Aesthetic clinics

An always-on AI SDR for aesthetic clinics.

Turns WhatsApp leads into booked consultations — and hands closers only the ones ready to buy.

525
booked consultations / month
<30s
first reply · any hour, any volume

The agent’s job is to bring twenty-five qualified people to a human closer. Not to close them.

The problem

The clinic — I’ll call them Concierge Doctor here — ran entirely on WhatsApp. When ten leads arrived at once, the tenth waited two hours. Aesthetic leads are impulse-driven — a two-hour reply is a dead lead.

The clinic answered about a quarter of the inbound it could generate, so it couldn’t buy traffic at all — more ads only made the queue longer. Bookings sat at five a month.

What I built

A WhatsApp concierge that qualifies, nurtures and follows up 24/7. Behind it, an operations cockpit that scores every lead, writes a brief for the human closer, and books the consultation.

Results — first month, one clinic
  • 500 leads handled
  • First response under 30 seconds, regardless of volume
  • Booked consultations 5 → 25 per month

At the clinic’s R$5,000 average ticket, twenty extra consultations is roughly R$100,000/month in new consultation pipeline.

Where I stopped the AI

The agent hands to a human on exactly three triggers: it fails to resolve an objection after three attempts, the lead asks for a person, or it detects it lacks the information to answer correctly.

The third one matters most. Most clinic bots fail by improvising through a knowledge gap — an invented price, a promised result. That costs trust, and it can cost a treatment. So the system recognises the edge of its own knowledge and stops there.

See it run

The same system, running on a made-up lead.

WhatsApp is just the transport. The qualification, scoring and closer-brief logic is the product — it runs the same on a chat, a form or a call transcript.

Demonstration — scripted flow, fictional data. Not a live system or real client data.
Concierge Doctor
online
DEMO
TODAY
Scripted demo — sit back
Operations cockpit live
Lead score
/100
Listening
Signals detected
Listening…
Closer brief

Booking
·
Booked Pending
Handed to a human closer with full context.
The actual solution

Screens from the system in production — the real interface behind the demo above.

Case study 02 · Customer support

Support that answers from the docs — and knows when it can’t.

A custom retrieval agent embedded in an ed-tech product. Instant, cited answers — and a clean escalation the moment it isn’t sure.

4h<5s
first response
71%
resolved without a human
90%+
accuracy · golden Q&A set

* Illustrative figures — final production numbers pending.

How it’s built
Knowledge
Document standardizationGolden Q&A setSemantic chunkingStructural chunking
Retrieval
Hybrid searchRerankingRRF fusionHyDE
Pipelines & eval
Automated ingestionRAG eval harnessLLM-as-judgeLatency & cost
Stack
LangGraphLangChainpgvectorQdrant

The agent answers what the docs already know — and routes everything else to a person, fast.

The problem

A growing ed-tech company — small team, thousands of teachers and students. Support was a few people answering the same questions across email and chat. The help centre was scattered, so the answer you got depended on who replied.

As sign-ups grew, response times slipped. Hiring more agents scales cost, not consistency — and the fastest-growing queue was the same fifty questions asked a hundred ways.

What I built

A custom retrieval-augmented agent, embedded in the product as a widget. It answers only from a curated, standardised knowledge base, always with citations, and escalates the moment retrieval confidence drops or the question is out of scope.

Results — first weeks in production
  • 71% of conversations resolved end-to-end without a human
  • 90%+ answer accuracy against a golden Q&A set
  • First response 4h → <5s
  • Support scaled with sign-ups, no new headcount
Where I stopped the AI

Every answer is grounded in an approved source and shows its citations. When the top retrieval score falls below threshold — or the question is out of scope, like pricing or anything account-specific — the agent doesn’t improvise. It escalates with the full conversation attached.

A confidently wrong support answer is worse than a slow one: it teaches the user something false and quietly erodes trust in the product.

See it answer

One question, traced end to end.

The widget is what the user sees. The trace beside it is the retrieval pipeline that produced the answer — and the guardrail that escalates instead of guessing.

Demonstration — scripted flow, fictional data. Not a live system or real client data.
Help Assistant
online
DEMO
Ask about the product…
RAG pipeline live
Query
Retrieval trace
Waiting for a query…
Answer confidence
Currently building · DecisionLog

Decision memory for architecture firms.

It turns the meetings that already happen into a searchable record of every decision — who made it, why, and what it changed.

Pre-launch · in pilot with souBIM, an architecture firm in Florianópolis

souBIM runs three or more projects at once. Every material change, structural revision and budget call happened in a Google Meet — and lived only in the head of Gabriela, the director. Nothing was written down.

Then Gabriela scheduled two to three months of maternity leave. Around fifty decisions per project would be made without her. It wasn’t a leave problem — it was a scaling problem: the firm couldn’t grow past one person’s memory.

How it works
Capture
Google Meet
Ingest
Tactiq webhook
Extract
LangGraph · Claude
Structure
Decision schema
Store
pgvector
Surface
Timeline · digest
Watch one decision get captured
Demonstration — scripted flow, fictional data. Not a live system or real client data.
Raw transcript · Residential Tower Alpha
Carlos (Structural) — For the 50-floor load I’d lock high-strength concrete, C50, on the main columns.
André (MEP) — Fine by me, doesn’t touch the risers.
Gabriela (Director) — Agreed. Let’s commit C50 for the main structural columns.
Lúcia (Architecture) — No objection here.
Extracted decision
✓ captured
Awaiting transcript…
Extracting decision
What the team sees

Every decision, filterable and attributed — surfaced as a project timeline and a “while you were gone” digest.

Milestone timeline
21 decisions
Meeting 19Email 1Decision 11Info 3Action 3
Estudo Preliminar
JAN 15 → FEB 15, 2026
InfoUpdated waterproofing spec received from supplier
SExternal·Feb 8
DecisionInstall LED lighting system with smart controls
GGabriela (PM)·Feb 7
DecisionAdd transfer beams at level 5 for the retail podium
CCarlos (Structural)·Feb 6
DecisionUse high-strength concrete (C50) for the main columns
CCarlos (Structural)·Feb 6
Anteprojeto
FEB 16 → MAR 31, 2026
Projeto Legal
APR 1 → MAY 15, 2026
Projeto ExecutivoToday
MAY 16 → AUG 31, 2026
Executive digest
21
Total decisions
6
Meetings
85%
Consensus
11
High impact
Key highlights
StructuralFeb 8, 2026
Updated waterproofing spec received from supplier
Supplier sent updated spec for foundation waterproofing.
Landscape
Explore living wall system for lobby atrium
Could create a signature design element and improve air quality.
Landscape
Preserve existing trees and install a green roof on the podium
Environmental benefits, community requirement, and LEED stormwater credits.
Electrical
Emergency backup power for residential floors
Discussion about generator sizing and UPS systems.
From the live product
Results — pilot
3mo → 30min
director catch-up · per project
95%+
decisions captured
~$25/mo
to run
  • 150–200 decisions a month captured with zero meeting friction
  • Meeting ends → decision logged in under 4 hours
Where I stopped the AI

The system extracts and structures decisions — it never makes or approves them. It records consensus and dissent instead of smoothing them into a clean answer, and it scores its own confidence on every extraction.

When a decision is low-consensus or contradicts an earlier one, it flags it for the director rather than asserting a tidy record. A confidently wrong decision log is worse than a flagged uncertain one — in a building, trust is load-bearing.

The system’s job is to give a director back her memory. Not to make her decisions.

More work

Other projects, across other worlds.

A few more at different stages — two shipping right now, and one that won a competition and then shipped to paying clients.

Finance · ed-tech In progress

AI Librarian for a financial-education platform

An in-product agent that answers from the platform’s videos and lessons, guiding members through their learning journey toward real results.

Pilot
Architecture · Upwork Delivered · ★★★★★

Project calls, turned into a searchable knowledge base

Ingested an architecture firm’s call recordings and turned them into a searchable content platform. Delivered on Upwork with a five-star client review.

“Deeply impressed with his knowledge of RAG systems, clear communication, and depth of architectural understanding.”

★★★★★Client feedback · 5.0 · Dec 2025
AI Knowledge Engine · RAG architecture
Creative · digital clone 1st of 200 · shipped

Digital clones of how a person decides and speaks

Chain-of-thought and few-shot systems with ElevenLabs voice, replicating a person’s decisions and style. First of 200 at Academia Lendária, then shipped to three paying clients.

5× interactions · 85%+ satisfaction
Selected experience

A decade of production engineering.

Independent, contract engagements — from cloud CRM and cross-border payments to the AI systems above.

SignalCore AI Enterprise AI Contract
AI Engineer · Autonomous AI decision systems for enterprise
2026 — present
The Sandbox Web3 gaming Contract
Backend Engineer · Blockchain virtual world · NFT ownership
2020 — 2021
dLocal Payments Contract
Backend Engineer · Cross-border payments infrastructure
2020 — 2021
XPO Logistics Logistics Contract
Backend Engineer · Freight transportation platform
2019 — 2020
Salesforce CRM / SaaS Contract
Backend Engineer · Cloud CRM platform
2016 — 2019
Recognition

Three wins in twelve months.

1st
Academia Lendária — style-replication AI, 200 participants
2025
1st
Ethereum Argentina (60+ teams) & Ethereum Uruguay (20+ teams)
2024
2nd
Chainlink Global Hackathon — on-chain insurance protocol View ↗
2024