Systems • Parallel Computing • ML • .NET/WPF • Clean Architecture
I build pragmatic, high-signal software: concurrent C++ systems, reliable data/ML pipelines, and desktop apps with strong UX. I optimize for clarity, performance, and reproducibility.
- Parallel & concurrent C++ (Intel TBB pipelines, work partitioning, lock-free patterns)
- ML prototypes (TensorFlow/Keras), dataset curation, metrics, and experiment tracking
- WPF/.NET with MVVM, testable layers, and accessible UI
- Strong foundations: algorithms, data structures, databases, CI/CD
| Project | What it does | Tech |
|---|---|---|
| genetic-timetable-scheduler | Genetic algorithm for course/classroom scheduling with constraint handling | Python, NumPy |
| parallel-web-scraper-tbb | C++20 scraper with TBB pipeline (fetch → parse → analyze → store), retries, auto-discovery | C++20, Intel TBB, libcurl, CMake |
| NASP-NoSQL-Engine | Educational KV store: MemTable → SSTable, compaction, LRU cache | Go |
| checkers-python | Checkers with Minimax + alpha-beta, GUI, transposition hints | Python, Pygame, SQLite |
| tweet-emotion-classifier | Tweet emotion multi-class classification with RNN baseline | Python, TensorFlow |
| solar-cycle-lstm | LSTM forecasting for sunspot time-series with robust loss | Python, Keras |
| medinsurance-ann-tensorflow | ANN regression for insurance costs with feature engineering | Python, TensorFlow |
| OOP2-Minotaur | Console labyrinth game: grid mechanics, fog of war, pickups | C++ |
Quick links
- 🔗 https://github.com/ljubogdan/genetic-timetable-scheduler
- 🔗 https://github.com/ljubogdan/parallel-web-scraper-tbb
- 🔗 https://github.com/ljubogdan/NASP-NoSQL-Engine
- 🔗 https://github.com/ljubogdan/checkers-python
- 🔗 https://github.com/ljubogdan/tweet-emotion-classifier
- 🔗 https://github.com/ljubogdan/solar-cycle-lstm
- 🔗 https://github.com/ljubogdan/medinsurance-ann-tensorflow
- 🔗 https://github.com/ljubogdan/OOP2-Minotaur
parallel-web-scraper-tbb
- Problem: crawling and analyzing large paginated catalogs efficiently.
- Approach: 4-stage TBB pipeline (download → parse → analyze → persist) with bounded concurrency and back-pressure.
- Highlights: retry policy, URL auto-discovery, metrics, CMake presets, clean separation of concerns.
genetic-timetable-scheduler
- Problem: hard constraints (rooms, clashes) + soft constraints (preferences).
- Approach: GA with tailored chromosome encoding, fitness, crossover/mutation operators.
- Highlights: constraint penalties, reproducible seeds, CSV exports and basic visualization.
NASP-NoSQL-Engine
- Problem: write-optimized KV store with predictable reads.
- Approach: MemTable (in-memory) → SSTables (immutable, on-disk) + compaction strategy.
- Highlights: append-only writes, index blocks, LRU read cache.
- Readable > clever: explicit dataflow, small modules, strong interfaces
- Determinism & reproducibility: pinned deps, seeds, scripts, CI-friendly outputs
- Observability: metrics, logs, assertions, fail-fast paths
- Performance last, but not never: profile before optimize; benchmark with baselines
- Docs as code: README-first, diagrams when helpful, examples runnable end-to-end
- Task-based parallelism and heterogeneous execution models
- Data-centric design for ML pipelines (lineage, data contracts)
- Strong MVVM patterns for WPF with testable UI logic
- Better ergonomics for C++ build systems and presets
- LinkedIn: https://www.linkedin.com/in/bogdan-ljubinković-61a758345
- GitHub: https://github.com/ljubogdan
If a project sparks your interest, open an issue or start a discussion.