Diego Souza · AI Engineer

Experience

Career history, in my own words. These are statements about my work, not results measured on this site.

English draft of the CV, pending the owner's review.

Career

  1. Mar 2025 to May 2026 · Juazeiro do Norte, CE

    AI Engineer, Grupo Intra de Ensino e Pesquisa a Distância

    • Designed and shipped to production a multi-agent WhatsApp assistant with Python, LangGraph, LangChain and MCP (Claude and OpenAI APIs). One agent classifies intent and sentiment and routes to WhatsApp Flows, tool use or a fallback; another writes the answer.
    • Connected it to REST APIs, webhooks and the Kommo CRM, automating the student journey in 4 steps (sign-up, enrollment, exams, certificates) for tens of thousands of students a month, with 20% less manual support.
    • Built a RAG pipeline with embeddings and semantic search on PostgreSQL (pgvector) over company documents, grounding answers in official sources.
    • Applied prompt evals, guardrails, hallucination mitigation, failure handling and human handoff, with daily CloudWatch monitoring and model cost control.
    • Led architecture and backend in Python and n8n, with Docker on AWS ECS, CI/CD on GitHub Actions and TDD; documented decisions in ADRs and adopted Spec-Driven Development with human review of code generated by coding agents.

    Stack: Python, LangGraph, LangChain, MCP, Claude API, OpenAI API, PostgreSQL, pgvector, Redis, MongoDB, Docker, AWS ECS, CloudWatch, GitHub Actions, NestJS, Go, n8n

  2. Feb 2024 to Jan 2025 · Fortaleza, CE

    Software Engineer, Yamaha Motor (via Compass UOL)

    • Built back-end services with NestJS, TypeScript and Python in microservices with Docker, JWT and OAuth, on Yamaha's e-commerce platform, with API caching.
    • Built Next.js and React features and contributed to Vibe, a community platform in an international project, with Scrum, code reviews and communication in English.

    Stack: NestJS, TypeScript, Python, Docker, Next.js, React, Scrum

  3. Aug 2023 to Feb 2024 · Juazeiro do Norte, CE

    AI Engineer (ICT Residency Fellow), Sociedade BRISA de Desenvolvimento de Tecnologia da Informação

    • In a team of 5 at the ICT residency (Softex and MCTI), built the Flask and FastAPI APIs and the React platform for lung disease diagnosis from images. The TensorFlow CNN reached 97.27% accuracy and was published at SBCAS.

    Stack: Python, FastAPI, Flask, React, TensorFlow, Oracle

  4. Jun 2023 to May 2024 · Juazeiro do Norte, CE

    AI Research Assistant, Universidade Federal do Cariri (UFCA)

    • Developed a low-compute method for knee osteoarthritis diagnosis in radiographs (VGG-16, PCA and Random Forest, 95.74% precision), published in the SEPEC proceedings.

    Stack: Python, scikit-learn, TensorFlow, PyTorch

  5. Jun 2022 to May 2023 · Cajazeiras, PB

    Software Engineer, Chama Inovação (startup)

    • Led, in direct contact with the client, a medical appointment platform: Node.js and Express backend with TypeScript, MongoDB and PostgreSQL, JWT and OAuth, real-time chat, payments and scheduling, and a React and Redux front end on AWS.

    Stack: Node.js, Express, TypeScript, React, MongoDB, PostgreSQL, Docker, AWS, Socket.IO, Scrum

Education

Publications

Certifications

Languages

Skills

Generative AI and LLMs
LLMs, prompt engineering, Claude API, OpenAI API, Amazon Bedrock, LiteLLM
RAG and vector stores
RAG, embeddings, semantic search, pgvector, PostgreSQL, Redis, MongoDB
AI agents
multi-agent systems, tool use, LangChain, LangGraph, MCP, human-in-the-loop, n8n
Python, APIs and backend
Python, FastAPI, REST APIs, webhooks, TypeScript, Node.js, NestJS, Go, microservices, JWT, OAuth
Cloud and DevOps
AWS (ECS, Lambda, EC2, S3, CloudWatch), Docker, GitHub Actions, CI/CD, TDD
AI evaluation and safety
prompt evals, guardrails, hallucination mitigation, Langfuse, LangSmith, OpenTelemetry, model cost control
AI in development
Claude Code, Cursor, GitHub Copilot, Spec-Driven Development, ADRs

Message me on WhatsApp

Plain-text résumé