AI Engineering · Machine Learning · Research
Computer Engineering student at UFPel and FAPERGS research fellow. I build applied AI workflows, evaluate learning systems, and turn research ideas into testable software.
eduardotbuss.github.io · GitHub · LinkedIn · ORCID · academic email
- Applied AI: Order Forge, built for a workshop challenge, turns customer PDF orders into validated EDIFACT. It combines deterministic parsing, optional local or API-based LLM extraction, catalog checks and human review. Its benchmark showed that catalog reconciliation, rather than reading the PDF, was the main bottleneck.
- Machine learning: I implement and evaluate neural networks, from a NumPy MNIST classifier to published CNN architecture experiments in PyTorch and TensorFlow.
- Research and harness engineering: I build quantum-fuzzy inference experiments and use reproducible inputs, baselines, tests and explicit failure cases to assess results.
At UFPel, my FAPERGS research with Tech&ApplieD-GM / Q-Flex covers quantum computing and fuzzy logic. My earlier ViTech work studied deep learning and computer vision. I am interested in AI and ML engineering opportunities where experiments become reliable systems.
| Project | What it is | Stack |
|---|---|---|
| Order Forge | PDF purchase orders to validated EDIFACT, with LLM extraction and human reconciliation. | Python, FastAPI, Next.js, Docker |
| Q-FIE | An interactive academic demo of fuzzy inference with classically simulated quantum circuits. | Python, Qiskit, FastAPI, React |
| Poker Decision Analytics | Analyzing 60,000+ poker hands to measure decision quality, variance and tail risk. | Python, Pandas, NumPy, Matplotlib |
| mnist-neural-network-numpy | A feedforward network with forward pass, backpropagation and gradients written out by hand in NumPy. | Python, NumPy |
| inventory-management-api | A multi-tenant inventory API where stock is derived from an append-only ledger instead of a mutable counter. | Python, FastAPI, SQLAlchemy, SQLite |
| quantum-algorithms-qiskit | Foundational quantum algorithms as executable notebooks: theory, circuit and measured result side by side. | Python, Qiskit, Jupyter |
- A Hybrid Classical–Quantum Formulation of Fuzzy CRI Inference with a Triage Case Study. IEEE CEC 2026 (WCCI), 2026.
- A Systematic Literature Review on Classical Data Encoding Strategies for Hybrid Quantum Machine Learning. ICEIS 2026, 2026. DOI
- QL-Implications: Quantum-Fuzzy Modeling and Application to Financial Decision-Making. ENIAC 2026, 2026 (accepted).
Python, PyTorch, TensorFlow, scikit-learn and NumPy for ML; Qiskit and PennyLane for research; FastAPI, SQLAlchemy, Docker and SQL for applications. I also use C and Java. For agent workflows, I work with structured handoffs, persistent context, test harnesses and human review.
Updated daily.
Project case studies, evidence and limitations: eduardotbuss.github.io