About

An extended third-person bio.

Lucas Rosenblatt is an Assistant Professor of Computer Science at Williams College whose work focuses on responsible AI/ML. His research develops methods for generating high-quality, privacy-preserving synthetic data, designs algorithms that reduce bias and promote fairness in machine learning, and improves our understanding of large language models in order to identify and mitigate systemic risks to users.

He completed his PhD in Computer Science at New York University, where he was a member of the Center for Responsible AI and the Theoretical Computer Science Group. He was supported by an NSF Graduate Research Fellowship. Before that, he worked at Microsoft as a ML Research Engineer for a couple of years. He completed his undergraduate degree from Brown University in 2019.

Research

I try to design and evaluate approaches that are theoretically sound, practically useful, and socially impactful:

  • Differentially private synthetic data: algorithms and reproducibility-focused benchmarks for deploying synthetic data in health and policy settings, including the question of whether “public” data is needed at all.
  • LLM privacy and evaluation: threat models for auxiliary-knowledge and fragment-inference attacks, private fine-tuning, watermarking, and bias-bounded evaluation of models used as judges.
  • Algorithmic fairness: revisiting foundational fairness metrics, characterizing when multiple group metrics can be approximately satisfied at once, and benchmarking uncertainty in fairness interventions.

A more complete list of my work is available on my Google Scholar profile.

Industry Research Experience

  • Google Research (NYC), Student Researcher, Summer 2025 - Winter 2026.
    Tabular data privacy problems with open-weight Gemma models.
  • Microsoft AI Development Acceleration Program (MAIDAP), ML Engineer/Researcher, 2019-2021.
    A rotational research program serving Microsoft organizations. Rotations: Grey Systems Lab, Microsoft + Harvard OpenDP (SmartNoise), Microsoft News.

Awards & Recognition

  • Communications of the ACM Research Highlight, 2026, SIGMOD Research Highlight, 2024, VLDB Best Paper, Runner-Up 2023, (Epistemic Parity: Reproducibility as an Evaluation Metric for Differential Privacy)
  • Theory and Practice of Differential Privacy (TPDP) 2025 Keynote
  • NSF Graduate Research Fellowship

Service

Reviewing. ICML (Gold Reviewer 2026, 2025), COLT 2026, TPDP 2026 & 2025, NeurIPS 2026 & 2025, ICLR 2026 & 2025, AISTATS 2026, 2025 & 2024, KDD 2025, FAccT 2025, SOSA 2024, CHI 2023.

Organizing.

Outside of Research

I converted a school bus into a mobile home, and I write whenever I can. I also make pots.