Senior Data Scientist & ML Engineer
Building open-source systems where clinical AI, scientific computing, and machine learning meet.
clinical AI · reinforcement learning · scientific ML · LLM tooling
I am Thomas Virdis, also known as CTCycle. I build open-source tools for clinical and scientific workflows, from medical computer vision and clinical copilots to reinforcement-learning experiments, computational chemistry, and local-LLM tooling.
My current work is as a Senior Data Scientist and ML Engineer at Inmatica S.p.A. My path combines bachelor's and master's studies in Biotechnology at the University of Genoa with a PhD in Engineering Sciences at the Vrije Universiteit Brussel. During my PhD, I worked on CheckPack, developing micro-sensors for real-time food-spoilage and quality monitoring.
Four projects that represent the main areas of my engineering work:
| Project | Focus | Stack |
|---|---|---|
| FAIRS Roulette Player | A local research workspace for DQN training and inference experiments, with dataset management, checkpoints, metrics, and session workflows. | PyTorch FastAPI React Tauri |
| XREPORT Radiological Reports | A client-server application for preparing datasets, training and validating models, and generating draft reports from X-ray images. | Medical AI Transformers FastAPI React |
| ADSMOD Adsorption Modeling | A scientific application for collecting adsorption data, fitting theoretical models, and training predictors from NIST and ARPA-E datasets. | Python RDKit NumPy Pandas |
| LLMeter Local Benchmarks | A Rust CLI for repeatable benchmarking of local OpenAI-compatible LLM providers, with performance metrics and exportable reports. | Rust LLM evaluation CLI |
- Building: Open-source ML tools for clinical research, medical imaging, reinforcement learning, computational chemistry, and LLM workflows.
- Learning: Front-end development and productionizing ML with Rust.
- Collaborating: On practical projects at the intersection of biotechnology, healthcare, and machine learning.
- Ask me about: Reinforcement learning, medical AI, or moving from biotechnology research into machine learning.
| Area | Working stack |
|---|---|
| Languages | Python, Rust, Java, JavaScript, SQL, Bash |
| ML and deep learning | PyTorch, TensorFlow, Transformers, Hugging Face, LangChain |
| Infrastructure | Docker, Kubernetes, Linux, REST APIs, PostgreSQL |
| Scientific computing | RDKit, NumPy/SciPy, Pandas, NIST databases, ARPA-E |
| Workflow | Git, GitHub Actions, Jupyter, VS Code, Postman, CI/CD |
- Now: Senior Data Scientist and ML Engineer at Inmatica S.p.A., working on applied AI tools for clinical and scientific workflows.
- Research: PhD in Engineering Sciences at the Vrije Universiteit Brussel, including the CheckPack micro-sensor project.
- Foundation: Bachelor's and master's studies in Biotechnology at the University of Genoa.
These cards show public-repository activity only. GitHub Actions refreshes the repository-owned SVGs daily, so the profile does not depend on a live public stats endpoint at render time.
The Trophy cards are generated from the maintained GitHub Profile Trophy implementation and refreshed by the same repository-owned workflow.
I am always interested in thoughtful collaborations involving biotechnology, healthcare, machine learning, and tools that help research move from experiment to usable software.