Marco Virgolin

About Me

Academic Insight. Industrial Impact.

Over 7 years in academia—culminating in a tenure-track position at the Dutch National Center for Math & CS—shaped my ability to think critically and creatively about complex problems. The past years in senior industry roles have expanded that foundation, from building production-grade ML systems to leading teams and guiding strategy. Today, I thrive at the intersection of research and application, bridging cutting-edge innovation with real-world implementation. From time to time, I still provide academic service.

Research Background

Business Impact

Hands-on Machine Learning

Leadership & Strategy

Machine Learning Deep Learning LLMs Responsible AI Generative AI Research & Innovation

My Journey

🤖 Head of AI

2024 - Present

InSilicoTrials Technologies (NL)

Built and led the AI team in the drug development-focused startup.
Collaborated with R&D, Engineering, and C-level to identify and pursue business value with AI.
Laid MLOps foundations: testing, CI/CD, pipelines, as well as onboarding materials.
Designed, developed/supervised, and deployed AI solutions covering supervised and unsupervised ML, explainable AI, synthetic data generation, multimodal info retrieval, and LLM-based multi-agent systems (incl. w/ MCP).
Contributed to large-scale grant and consulting projects.
Academic collaborations such as PhD supervision and teaching.

📊 Senior Data Scientist

2023 - 2024

INGKA Digital [IKEA] (NL)

Conducted R&D on generative AI, focusing on mechanistic intervention on LLMs and diffusion models.
Developed semantic search and generative AI prototypes.
Advised on responsible AI implementation strategies.
Conducted technical interviews for senior and principal roles.
Mentored less senior colleagues.

👨‍🏫 Scientific Staff Member (Tenure Track)

2021 - 2023

Dutch National Center of Math & Computer Science [CWI] (NL)

Research on evolutionary algorithms, explainable AI, deep learning.
Published at top-venues, incl. ICML, NeurIPS, GECCO.
Developed open-source algorithms.
Co-authored research proposals with industry partners.
Supervised 4 PhD students.
Academic service.

🔬 Postdoctoral Researcher x2

2019 – 2021

TU Chalmers (SE)

Worked on explainability in NLP, active learning, robotics, and human-AI interaction.
Supervision of M.Sc. students and education.
Academic service.

2019 – 2020

Dutch National Center of Math & CS + TU Delft (NL)

Research on automatic discovery of deep learning architectures.
Facial emotion recognition using deep contrastive learning.
Supervision of M.Sc. students and education.
Academic service.

🎓 Ph.D. in Evolutionary Machine Learning

2015 - 2020

National Math & CS Center + TU Delft, NL

Researched evolutionary algorithms grounded in information theory.
Developed algorithms to discover interpretable symbolic regression models.
Applied research to pediatric radiotherapy.
Won several awards.

Honors and Awards

🏆 Awarded SIGEVO Best Ph.D. Dissertation (2020), HUMIES Silver award (2021), and 2×Best paper awards.
🇪🇺 Served as external expert for the European Commission on research calls (25M €, 2024).
📝 Served in the program committee of several conferences and workshops: GECCO, ECML, PPSN, and more.
🕵️‍♂️ Served as reviewer for several journals: Machine Learning, IEEE TEvC, Soft Computing, and more.
🤺 Co-organizer of the Symbolic Regression competition 2022 and 2023, where I co-designed novel benchmarks and evaluations.
🇸🇪 Recipient as co-applicant of a 300,000 SEK grant by Area of Advance Health Engineering, TU Chalmers (2021).
💻 Recipient of NWO SURF computing grants and ACM student travel grants (during Ph.D.).
👨‍🏫 Honored to give invited lectures/talks at multiple venues, including: UvA, TU Delft, TU Chalmers, MIT, G-Research.

Selected Projects

multi-agent system using model context protocol (MCP) to perform tasks related to drug development

🦙 LLM Multi-Agents

Led the development and contributed hands-on to an LLM-based multi-agent system capable of handling different types of queries by autonomously operating tools. The architecture featured a Model Context Protocol (MCP) server coded in-house to serve several specialized tools. These included retrieval augmented generation (RAG) with sources both internal or external to the company, and functionalities of the simulation platform within which the multi-agent system was deployed, thus offering seamless integration.

LLMs RAG Llama-index MCP Multi-agent system

🔍 Explainable AI Algorithms

Contributed first-hand to explainable AI with: (1) An award-winning algorithm and respective publications to discover mathematical equations from data (symbolic regression). Check out: SRBench at NeurIPS | SIGEVO's Best PhD Dissertation Award 2021 | Silver Award at 2021 Humies Competition | Open source repo. (2) A simple but effective algorithm to explain black-box ML models by counterfactual explanations: "how should the input change to obtain a different output?". Check out: Publication on Artificial Intelligence (Elsevier, open access) | Open source repo | Colab example.

Explainable AI Symbolic regression Black-box AI Counterfactual explanations Genetic algorithms

💊 Responsible AI in Pharma

Worked on bringing into practice and contributed to the discussion on responsible AI, such as FDA's guidance proposal on the use of AI for drugs (comment on behalf of the company). Further, led the development of a technical and actionable protocol for quantifying the level of fidelity, utility, and privacy of Generative AI-based synthetic data in healthcare. Measuring privacy risks in particular required mapping regulatory guidelines from the EU's GDPR and European Medicines Agency into appropriate metrics and simulated cyber-attacks from academic literature on synthetic data and anonymization.

Responsible AI Regulations Generative AI Synthetic data Digital twins

Bridging Research and Industry in AI