Skip to content
View dheiver2's full-sized avatar

Block or report dheiver2

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
dheiver2/README.md
Typing SVG

Website LinkedIn X Medium Google Scholar

Profile Views


👋 About Me

  • 🇧🇷 Brazilian AI scientist and engineer with 15+ years of experience in Artificial Intelligence, MLOps, AIOps and Data Science.
  • 🎓 Doctorate in Artificial Intelligence, focused on distributed systems, generative AI and autonomous orchestration.
  • 🥭 Creator of Mangaba AI, a growing ecosystem of self-improving, multi-platform autonomous agents.
  • 🚀 Designs production-grade AI pipelines and intelligent automation for real-world use cases across health, retail and public services.
  • ✍️ Writes and publishes technical content on modern AI engineering practices.
  • 🌱 Currently exploring neuro-symbolic AI, self-adaptive AIOps frameworks and applied AI ethics.
  • 🤝 Open-source advocate and mentor for emerging Brazilian tech talent.

🔬 Measured on the bench

Latest results from my own experiments, updated automatically from the commits where they were measured.

  • In 6 rounds across a set of 33 difficult tasks, the cascade achieved an average of 26.8 correct answers against 25.7 for Claude: a small difference that cannot yet be separated from the noise between rounds. → commit
  • When extracting and running JavaScript expressions directly from the question text (without calling the model), the median time (p50) dropped from 22.4s to 11.7s in specific tests. → commit
  • In the comparison across 30 difficult tasks over two rounds, cascade v4 got 49 answers correct and Claude 48, resulting in a technical tie. → commit

Weekly notes (pt-BR): Notas de Bancada · dheiverphd.com


🚀 Featured Projects


📊 GitHub Analytics


🏆 Trophies


🛠️ Tech Stack

Languages

Python R Julia Go Rust TypeScript Swift

AI & Machine Learning

TensorFlow PyTorch scikit-learn Hugging Face LangChain OpenCV

MLOps & AIOps

Docker Kubernetes MLflow Airflow Grafana Weights & Biases

Cloud & Infrastructure

AWS Azure GCP Terraform

Data & Analytics

Pandas NumPy Apache Spark


📫 Connect with Me

Website Email LinkedIn X Medium Google Scholar

Popular repositories Loading

  1. mangaba-claude-skills mangaba-claude-skills Public

    Skills Claude prontas para PMEs brasileiras: Financeiro, Marketing, Atendimento e RH — com Scheduled Tasks e Artefatos

    3 1

  2. mangaba-agent mangaba-agent Public

    🥭 O agente de IA que cresce com você — auto-aperfeiçoável, memória persistente, multi-plataforma

    Python 3

  3. tessera tessera Public

    A harness for coding agents: runs Claude Code, Codex, OpenClaw and Hermes on the same task and keeps the answer that passes your tests. Gateway-agnostic, zero dependencies.

    JavaScript 3

  4. gmail-to-sheets gmail-to-sheets Public

    JavaScript 2

  5. PDF2SheetBot PDF2SheetBot Public

    Python 2 1

  6. wifi-sensing wifi-sensing Public

    Análise de sinais Wi-Fi com IA para sensoriamento ambiental device-free (detecção de movimento/presença via RSSI), com dashboard PySide6 e interpretação por LLM local (BitNet).

    Python 2 1