About

Recent advances in LLMs, foundation models, and AI agents have rapidly expanded the role of AI in science. These systems are now being used to read literature, generate hypotheses, write and execute code, design molecules and experiments, interact with simulators, and assist in scientific reasoning. As AI systems become increasingly embedded in scientific workflows, the key question is no longer only what can they do, but how should we build and use them reliably for scientific discovery.

The AI Scientist Summer Workshop brings together researchers across AI and science to discuss the emerging vision of AI systems as scientific collaborators. We aim to examine both the opportunities and limitations of current models, and to ask what scientific problems are best suited for AI agents and foundation models.

— Central themes

Reliability & verification

Producing scientific outputs that are trustworthy and independently checkable.

Evaluation & reproducibility

Benchmarks and protocols for assessing scientific reasoning and reproducing results.

Closed-loop discovery systems

Hypothesis generation, experimental design, and closed-loop autonomous experimentation.

Foundation models for science

Models for molecules, sequences, simulations, and multi-modal scientific data.

Human–AI collaboration

How scientists and AI systems divide work, build trust, and reason together.

Scientific applications

Domains where AI agents and foundation models drive new discovery — biology, chemistry, materials, physics, and beyond.

Confirmed speakers
Marinka Zitnik
Harvard Medical School
Yuanqi Du
Microsoft Research
Nathan Frey
Anthropic
Schedule
08:00 – 09:00
Registration
09:00 – 09:10
OpenOpening remarks
09:10 – 09:40
Invited TalkNathan FreyAnthropicTeaching AI Scientific Taste and Judgement
09:40 – 10:10
Invited TalkVivek NatarajanGoogle DeepMindAdvancing Science and Medicine with Collaborative AI Agents
10:10 – 10:20
Contributed TalkSina BarazandehAgentic Fault Recovery for Autonomous Orchestration of Programmable Cloud Laboratories
10:20 – 10:40
BreakCoffee break
10:40 – 11:10
Invited TalkYuanqi DuMicrosoft ResearchTowards Generalist Agents for Accelerating Scientific Discovery
11:10 – 11:40
Invited TalkMarkus BuehlerMITAdaptive Swarms Across Scales
11:40 – 11:50
Contributed TalkRon ArelAutomating Discovery by Focusing on the Nature of the ‘Loop’
11:50 – 13:20
Lunch, poster & social
13:20 – 13:50
Invited TalkMarinka ZitnikHarvard Medical SchoolEmpowering Biomedical Discovery with AI Scientists
13:50 – 14:20
Invited TalkRafael Gómez-BombarelliMIT, Lila SciencesToward a Shared Representation of Matter: Alignment, Scaling, and the Path to Scientific Superintelligence
14:20 – 14:30
Contributed TalkTianyu Liu, Allen WangBenchmarking AI Agents for Addressing Scientific Challenges Across Scales
14:30 – 15:20
BreakCoffee break
15:20 – 15:50
Invited TalkBodhisattwa MajumderAI2Intriguing Data-driven Speculations with AI on the Possibilities in Science
15:50 – 16:20
Invited TalkPaul LiangMITSelf-Evolving AI for Open-Ended Discovery
16:20 – 16:30
Contributed TalkHarit Vishwakarma, Klara KalebPaperena: An Environment for Continual Evaluation and Improvement of AI Scientists
16:30 – 16:40
CloseClosing remarks
16:40 – 18:00
Poster session & social
18:30 – 21:30
Post-workshop social · AI-Bio Meetup in BostonCambridgeSide, Cambridge · RSVP on Luma
Posters
Morning Session26 posters
Adaptive Exploration–Exploitation Control for Agentic De Novo Drug Design
Jinyeop Song
Agentification of multi-omics pipeline and analytics for drug discovery applications
Giorgio Gaglia
AI CFD Scientist: Toward Open-Ended Computational Fluid Dynamics Discovery with Physics-Aware AI Agents
Shaowu Pan
AI-research assistant for experimental quantum science
Evgeny Redekop
AutoScientists: Self-Organizing Agent Teams for Long-Running Scientific Experimentation
Ada Fang
Calibratable Generative Agents: Bayesian Inference for Behavioral Epidemics
Khondoker Nabi
Chorus: An Agentic Framework for Comparative Regulatory Genomics with Sequence-to-Function Oracles
Luca Pinello
Document-as-Image Representations Fall Short for Scientific Retrieval
Ghazal Khalighinejad
Evidence Before Hypothesis: Human-AI Collaboration in Single-Cell RNA-seq Discovery
Ishan Khurjekar
FEM-Bench: A Structured Scientific Reasoning Benchmark for Evaluating Code-Generating LLMs
Erfan Hamdi
jx: an agent-composable marimo notebook catalog for scientific data exploration
Shantanu Singh
Leading multi-agent teams for scientific discovery
Ilknur Icke
LLM Agents for Hypothesis Generation and Experimental Design in Heterogeneous Catalysis
Anna Kelmanson
Medea: An AI agent for therapeutic reasoning across biological contexts
Michelle M. Li
Multi-agent reasoning for exploring the protein universe
Blake Lash
Multimodal AI and Participatory Governance for Older Adult Pedestrian Safety Decisions
Alex Quistberg
Observe or Perturb? The Data Mix That Reaches General Pathway Identifiability with the Fewest Total Samples (A Toy Model)
Elizabeth Hudson
Optimizing Behavioral Experiment Design using AI Simulation
Sanchaita Hazra
SAGE: Scalable Agentic Grounded Evaluation for Crop Disease Diagnosis
Tirtho Roy
Scaling real-world environments for data-driven discovery
Hanane Nour Moussa
SIA: Self Improving Agent with Harness and Weight Update for Scientific Research
Vignesh Baskaran
Terminal-Bench Science: Evaluating AI Agents on Real Computational Workflows across the Natural Sciences
Steven Dillmann
Therapy Agent: A G2P Platform-Traversing Agent for Target Identification
Amit Shenoy
Toward an AI Scientist for Functional Genomics: Coupling STRAND with Agentic In-Silico Perturbation Screening
Boyang Fu
Verification and Validation of Open Source AI-enabled Biomedical Tooling at Scale
Nicholas Alico
Where Agents Accelerate Small Molecular Discovery in Practice
Nate Gruver
Afternoon Session26 posters
Accurate Simulation and LLM-Enabled Orchestration for Fault-Tolerant Programmable Cloud Laboratories
Sina Barazandeh
Adaptive Knowledge Graphs with Scaling-Law-Based Completeness Assessment for Trustworthy AI in Self-Driving Materials Laboratories
Jing Luo
Agentic Calibration of Superconducting Qubit Devices
Beatriz Yankelevich
Agentic Discovery of Exchange-Correlation Density Functionals
Titouan Duston
AI Co-Pilot for Cancer Research
Anas Zafar
AI predictions and the expansion of scientific frontiers: Evidence from structural biology
Mengyi Sun
ALEMBIC: Precise chemical reaction extraction from scientific PDF documents
Islam Tayeb
Auditing Foundation Models for Scientific Decision-Making: Causal Safety Evaluation in Clinical Time Series
Aditya Kumar Karna
Can AI Scientists Discover Better Drugs than Human?
Xiwei Cheng
Can LLM Agents Do Research on Complex Systems? An Introduction to EpidemIQs
Mohammad Hossein Samaei
Collider-Bench: Benchmarking AI Agents with Particle Physics Analysis Reproduction
Darius A. Faroughy
Curie: Accelerating Oncology Discovery with AI Assistants
Tom Galeazzo
DREAMS OER: A Trustworthy Hierarchical Multi-Agent Framework for Resource-Aware, Hypothesis-Driven Acidic OER Catalyst Discovery
Ziqi Wang
Harnessing AtomisticSkills for Agentic Atomistic Research
Bowen Deng
Learned Hypothesis Generation for Automatic Scientific Discovery
Reece Adamson
LLM-driven model discovery for earthquake forecasting using a decade of prospective forecasting experiments in California
Sam Stockman
Paperena: An Environment for Continual Evaluation and Improvement of AI Scientists
Harit Vishwakarma
Planning What to Know:Evidence Coverage Tracking for Tool-augmented Biomedical Agents
Xinyi Xie
Protein Design with Agent Rosetta: A Case Study for Specialized Scientific Agents
Jacopo Teneggi
Regulatory Sandboxes as Testbeds for Trustworthy AI Scientists in Healthcare
Gokhan Ertaylan
RxFM: A Multimodal Foundation Model for ALS Therapeutic Science and Drug Discovery
Rohan Bhukar
Spatial Dissection of the Triple Negative Breast Cancer Tumor Microenvironment Using CODEX Multiplex Imaging.
Nidhi Kadam
VERITAS: Verifiable Epistemic Reasoning for Image-Derived Hypothesis Testing via Agentic Systems
Lucas Stoffl
Voxels to Vectors: Making Medical AI Explainable
Yashbir Singh
Why Learned PDE Solvers Explode: Semigroup Transformers for Stable Long-Horizon Dynamics
Dongzhe Zheng
ZooBench: Evaluating vision-language models on real citizen-science tasks
Nolan Koblischke
Post-Workshop Social

Join us for the official post-workshop social, the AI-Bio Meetup in Boston. We're opening the evening to the broader Boston AI + Bio community for food, drinks, and conversation. Come reconnect with old friends, meet new collaborators, and continue the conversations from the day. You don't need to have attended the workshop to join.

Location: CambridgeSide, 100 Cambridgeside Pl, Cambridge, MA 02141

RSVP on Luma →

Registration

The workshop is in person, with seats limited to 100 attendees. Register through the form below to attend.

Contributed talks & posters: If you would like to present a contributed talk or poster, please submit your proposal directly through the registration form. There is no separate submission portal.

Open registration form →

Organizers
Ada Fang
Harvard University
Wengong Jin
Northeastern University
Yuanqi Du
Microsoft Research
Student helpers
Yikun Zhang
Northeastern University
Xiwei Cheng
Northeastern University
Botao Yu
Ohio State University
Yasha Ektefaie
Broad Institute