~ ❯ whoami

renan mello

AI Engineer

I build production LLM agents for customer support on voice (IVR) and WhatsApp, with TypeScript/Node.js on AWS (Amazon Bedrock, Lambda).

Open to remote roles · Based in Brazil (UTC-3), overlapping with European afternoons and US East Coast hours.

Download Resume

~15%fewer agent misroutes after fixing LLM failure patterns

60,000COP30 attendees served by the SECOP housing system

70%+faster listing corrections for real estate agencies

Thoughtful interfaces. Practical engineering.view projects

01 cat about.md

I’m a full-stack developer based in Belém, Pará, Brazil. My experience spans financial systems, web applications and backend engineering for AI agents.

I’m passionate about Web3 and the financial market, with a focus on delivering a good, intuitive user experience.

More about my background

02 git log --author=renan

The path so far.

2018 — present · 6 commits

deeb2dbHEAD → now

September 2025 - Present

Accenture - Custom Software Engineering Analyst

Belém, Brazil

Build conversational AI agents with generative AI (LLMs on Amazon Bedrock) for a large company's customer support in 6 Latin American countries, on voice (IVR) and WhatsApp channels.
Developed AI outbound voice campaigns covering product offers, objection handling and consent confirmation.
Implemented self-service use cases integrated with corporate APIs, and tuned routing between the intent classifier, use cases and human handoff, including contact logging for support KPIs.
Designed prompts and deterministic flows (static responses, forced tool calls, variable capture) that reduced LLM hallucinations and kept answers aligned with business-approved text. Fixed failure patterns that caused wrong use case execution, cutting agent misrouting and response errors by ~15%.
Ran root cause analysis on QA and production incidents, investigating CloudWatch and OpenSearch logs and confirming or ruling out hypotheses through local reproduction.

TypeScript
Node.js
Amazon Bedrock
AWS Lambda
DynamoDB
S3
EventBridge
OpenSearch
CloudWatch
Jenkins

ca5a4a2

February 2025 - November 2025

Contract - Fullstack Developer

SECOP - Secretary of COP 30 - Belém, PA, Brazil

Developed a real estate management system for SECOP (Extraordinary Secretary for COP30, PA, Brazil) to support infrastructure for COP30, expected to host over 60,000 attendees.
As Belém has only ~30,000 hotel rooms, the system enables the secretary to onboard real estate agencies, allowing them to list private properties for short-term rental to meet the remaining demand.
Streamlined the property onboarding process for real estate agencies, significantly accelerating the standardization and validation of listings to match COP30 requirements, cutting correction and response time by over 70%.

TypeScript
Next.js
Tailwind CSS
PostgreSQL
AWS

git log --all 2024 — 2018 · 4 roles

f0cdba9

August 2022 - August 2024

XP Investimentos - Investment Advisor

Biggest broker in Brazil

Assisted clients with their investment portfolios at one of the oldest XP Investimentos offices, overseeing more than 1 billion Reais in assets under custody.
Developed Python automation tools to help other advisors improve the speed and accuracy of investment visualizations for their clients, significantly improving quality and optimizing time. Previously, this was a manual Excel process that consumed a considerable amount of the commercial team's time.

Python

b655c01

January - August 2021

INOA - Fullstack

Rio de Janeiro B2B company with about 50 software developers that work with the biggest financial market brokers in Brazil. The company's main goal is to automate, manage and integrate processes, in order to create a more efficient financial market.

Implemented automatic execution of market options on their maturity date, directing the funds to respective wallets, and developed a dedicated client page. This involves retrieving public data from the B3 (Brazilian Stock Exchange) using a web scraping and aligning it with the funds' wallets.
Created scripts for resetting and applying maintenance changes on the client server.
The development was carried out using Python, Django, SQL, and JavaScript.

Python
Django
JavaScript

c25cabf

July 2020 - January 2021

AGE PA - IT Manager

State-owned audit firm.

Managed the infra net server as well as the firewall.

Windows Server

dfdbc74

November 2018 - December 2019

PRODEPA - Intern

Set up and installed network switches across Pará state using switching programming language, including public schools and government agencies.

Monitored the network using Zabbix to ensure prompt action in case of failure, guaranteeing uninterrupted internet service throughout the state of Pará.

Switch Programming

03 ls -la ./projects

total 2 · featured

drwxr-xr-x
01

Brigabis Botin production

WhatsApp assistant running in production for a confectionery business. Uses retrieval-augmented generation (RAG) to answer customer questions, plus an AI configuration agent that proposes changes to the bot's content and settings, applied only after human review.

TypeScript
Node.js
Express
PostgreSQL
Next.js
AI Agents
RAG

drwxr-xr-x
02

AI Sports Coachshipped

A tool that analyzes running data and provides insights into performance and training.

Python
React Native
FastAPI
Machine Learning
AI
TypeScript

04 cat skills.yml

5 groups

AI & LLMs: Amazon Bedrock
prompt engineering
agents and tool calling
agent orchestration
RAG
deterministic conversation flows
voice (IVR) and WhatsApp channels
Languages: TypeScript
Python
JavaScript
SQL
Frameworks: Node.js
Next.js
React
React Native
FastAPI
Django
Express
Cloud & DevOps: AWS (Lambda, DynamoDB, S3, SQS, EventBridge, OpenSearch, CloudWatch)
Docker
Git
Jenkins
Databases: PostgreSQL
DynamoDB
MongoDB
OpenSearch
Languages (spoken): Portuguese (native) · English (C1)