I build applied AI software in Python and C++, with a focus on computer vision, native model inference, and LLM-based applications. I study Artificial Intelligence at Johannes Kepler University Linz and am interested in bringing AI into practical software and industrial systems.
SmartSpray Vision Controller connects crop/weed detection to a deterministic eight-channel spray-command simulation.
The project combines Python model fine-tuning and evaluation with native C++17 inference using OpenCV and ONNX Runtime. The native application runs without Python, PyTorch, or a GPU.
Engineering work includes model evaluation, Python/C++ inference-consistency checks, explicit input contracts, automated tests, and versioned model artifacts. A saved demonstration shows predictions, selected targets, and the resulting command schedule.
Camera geometry and actuation are simulated; this is an independent portfolio project, not a field-deployed or real-time control system.
Explore the project · Release v0.1.0
- Applied AI and computer vision applications in Python and C++
- Native model inference, evaluation, and integration into software systems
- Clear interfaces, automated testing, and reproducible workflows
- Tender qualification — private client project: built a modular Python pipeline for PDF, DOCX, and XLSX ingestion, OCR fallback, LLM-assisted classification and structured extraction, domain decision rules, and Excel/JSON outputs. Uses typed interfaces, pytest, Ruff, Pyright, and pre-commit.
- AI sales assistant — private collaborative project: contributed to an open, unmerged development branch covering aiogram/LangGraph routing, RAG retrieval, async RetailCRM integration, human handoff, and tests.
- scikit-llm #118 — Anthropic API support: added provider and credential integration across multiple model families, with tests; merged upstream.
- scikit-llm #124 — model constants: centralized model constants across provider modules; merged upstream.
- Transport Catalogue — Yandex.Practicum C++ course project with JSON I/O, graph routing, SVG rendering, and Protocol Buffers serialization.
- Spreadsheet — Yandex.Practicum C++ course project with formula parsing, dependency tracking, cycle detection, and recalculation caching.
- Search Server — Yandex.Practicum C++ course project with TF-IDF ranking and parallel processing using Intel TBB.
- Languages: Python, C++17.
- Applied AI / Computer Vision: PyTorch, Ultralytics YOLO, OpenCV, ONNX, ONNX Runtime, object detection, model evaluation.
- Software engineering: CMake, CTest, Python unittest, pytest, typed Python, Ruff/Pyright, pre-commit, Git, Linux; Protocol Buffers, graph algorithms, parsing, and parallel algorithms. SmartSpray uses Python unittest and CTest; pytest belongs to earlier projects.
- Applications / integrations: LLM APIs, LangGraph, async HTTP, external APIs, SQLite, PDF/DOCX/XLSX processing.