🚀 Full Stack Developer building AI-powered applications and orchestrating autonomous coding agents in production.

I don't just use AI to help me code — I work AI-natively. Spec-driven development executed by coding agents (Claude Code), with LLM-as-Judge pipelines where independent models cross-validate every output before it ships. If the logic isn't in the spec, it doesn't exist.

⚙️ How I work:

→ Spec-Driven & Detailed — rigorous technical design before any code is written. Bounded contexts, idempotent operations, decoupled services.

→ AI-Powered Implementation — coding agents execute. I steer the model and own architectural oversight, feedback loops, and quality control.

→ LLM-as-Judge Quality Gates — every output runs through independent LLM review pipelines that cross-validate against the spec.

→ Autonomous Agent Orchestration — designing self-correcting architectures where specialized agents handle infra provisioning, implementation, review, and production monitoring.

📊 What I build: an LLM-powered platform that streamlines audit reporting and compliance analysis end-to-end, helping public-sector teams move from manual workflows to AI-native operations at scale.

🎓 I also teach. I lead workshops and train non-technical professionals — auditors, public servants, and decision-makers — on how to apply AI to real audit work: prompt engineering, LLM tooling, and practical AI-native workflows that hold up under government-grade scrutiny.

🐝 Currently exploring swarms of specialized coding agents for end-to-end feature delivery — from infra provisioning on AWS / GCP to self-correction loops running in production.

🛠 Tech I work with:

• AI & Agents: Claude Code, LangChain, OpenAI, HuggingFace, RAG pipelines, LLM-as-Judge, prompt engineering, skills, autonomous agent orchestration, spec-driven development
• Backend & Data: Python, FastAPI, Pandas, PyTorch, Transformers, SQL, PostgreSQL, SQLite, Redis
• Frontend: React, Micro-frontends, Streamlit
• Cloud & Infra: AWS, GCP, Azure, Docker, Linux, serverless / Lambda-style functions
• Audit & Analysis: Fraud detection, pattern recognition, anomaly detection, machine learning, cross-referencing datasets
• Leadership: Team leadership, project management, architectural instincts, pragmatic & data-driven decision making

Rocha, A. L. M., Correa, T.P., CGU-Insight: Artificial intelligence at the service of internal auditing (2025). https://revista.cgu.gov.br/Cadernos_CGU/article/view/869

Paiva, E., Pereira, F., Carvalho, D., Junior, N., Oliveira, R., Bonifácio, S., Rocha, A., Oliveira, H., Cezar, F., & Junior, H. (2024). Continued pre-training of LLMs for Portuguese and Government domain: A proposal for product identification in textual purchase descriptions. Proceedings of the Association for the Advancement of Artificial Intelligence (AAAI). https://openreview.net/pdf?id=HBDb1ybEcs

Rocha, A. L. M., Rezende, M. S., & Oliveira, T. C. (2022). Alice: Desafios, resultados e perspectivas da ferramenta de auditoria contínua de compras públicas governamentais com uso de inteligência artificial. Revista da CGU, 14(26), 296-308. https://doi.org/10.36428/revistadacgu.v14i26.530

• Twitter: @andrecorumba
• LinkedIn: in/andrecorumba/