💫 Software Engineer with a passion for Language Models, open source and knowledge sharing.

🔍 Previously at 01S, I specialized in information extraction and retrieval from unstructured documents, making valuable information accessible to Italian citizens.

👨‍💻 Now at deepset, I contribute to Haystack, an open-source LLM framework and its ecosystem. I enjoy engaging with the community and sharing what I learn and spend time working on.


📬 stefanofiorucci (at) gmail (dot) com

---

Languages: Python, Java
LM inference/training: Hugging Face Transformers/TRL, Sentence Transformers, Ollama, vLLM
ML: scikit-learn, fundamentals of PyTorch
Databases: Solr, pgvector, Pinecone, Qdrant
Information extraction: Scrapy, Tika, Tesseract, Unstructured
Visualization: Gradio, Streamlit
Deployment/DevOps: Docker, FastAPI, GitHub Actions

Months ago I published my first Reinforcement Learning experiment: 
👑 🗓️ Qwen Scheduler GRPO

Using GRPO, I taught a model to create a schedule from a list of events and priorities.

The repo just hit 256 stars ⭐🎉
Time to celebrate!

Take a look if you are interested in Reward Function Design, Reward Hacking and applied RL stuff.
🔗 Links below.

More experiments coming soon...

Stefano Fiorucci

AI/Software Engineer | Open Source | Haystack LLM Framework 🏗️ maintainer

Train a Language Model with GRPO to create a schedule from a list of events and priorities

taplio

Stefano Fiorucci's Linkedin Analytics