Projects from the Veri Bilimi Okulu (VBO) AI@LLM Bootcamp (2026): seven weekly AI engineering projects, from fine-tuning a transformer to an MCP-based SQL agent.
Built by a Senior Data Scientist (11+ years in finance and e-commerce) who builds production-ready AI systems end to end.
| Week | Folder | What it does |
|---|---|---|
| 1 | week1-text-classification |
Fine-tunes DistilBERT (PyTorch + Hugging Face) for IMDb sentiment classification. |
| 2 | week2-encoding |
Turkish sentiment analysis: integer, one-hot and TF-IDF encodings with classical models vs. transformer embeddings, zero-shot classification and Turkish BERT fine-tuning. |
| 3 | week3-embedding |
Compares Turkish embedding models on similarity, analogies and clustering. |
| 4 | week4-vectorization |
Hybrid search over a 78-book library: BM25 + vector search combined with Reciprocal Rank Fusion. |
| 5 | week5-structured-output |
Schema-validated LLM extraction with Pydantic, retries and fail-fast error logging. |
| 6 | week6-rag |
HR chatbot with memory and source citations (LangGraph checkpointing + Chroma), comparing hosted and local models. |
| 7 | week7-mcp |
FastMCP SQL agent with five tools and a query guard. |
Each folder has its own README with setup steps and results.