avatar

Defne Circi

Ph.D. Student
Duke University
defne.circi@duke.edu


About Me

Currently, I am an AI Resident at Lila Sciences and a Ph.D. candidate at Duke University. I am very fortunate to be advised by Prof. Cate Brinson. Prior to starting my PhD, I was an undergraduate student in materials science and computer science at Sabanci University.

Research Interests

I am interested in multimodal AI systems for scientific reasoning and knowledge extraction, with a focus on autonomously interpreting diverse scientific data sources including charts, tables, figures, and text to structure knowledge in machine-understandable form. My current work spans developing and evaluating large language models and vision-language models for interpretable information extraction from scientific literature, as well as building tool-augmented agentic workflows for verifiable scientific reasoning. My goal is to unify these capabilities into scalable pipelines that power accelerated discovery in science.

If you’re interested in discussing research ideas or just want to chat, please contact me via email!

News

  • [Apr 2026] Co-organized the AI4MAT Workshop at ICLR 2026 in Rio de Janeiro, Brazil 🇧🇷.
  • [Mar 2026] Awarded the 2026–2027 Maclin Community Connections Grant to support organizing the 3rd Annual LLM Hackathon for Applications in Materials & Chemistry at Duke.
  • [Feb 2026] Started my AI Residency at Lila Sciences in Boston, MA.
  • [Dec 2025] Gave an invited talk titled "Scientific Reasoning and Information Extraction with Language Models" at Middle East Technical University (METU).
  • [Dec 2025] Presented Verifiable Chemical Reasoning through Tool-Calling Agentic Workflow at Scaling Environments for Agents Workshop at NeurIPS 2025 in San Diego!
  • [May 2025] Started my internship at IBM research in San Jose, California.
  • [Dec 2024] Presented 'Information extraction from diverse materials science charts/plots' at MRS in Boston.
  • [Dec 2024] Attended EMNLP 2024 in Miami.
  • [May 2024] Won an Anthropic Award in 2024 LLMs for Materials Hackathon with the project MaSTeA, an interactive web app that enables easy testing of LLMs to evaluate their strengths and weaknesses across various subfields of materials science.
  • [Apr 2024] Presented 'Using Large Language Models for Data Extraction from Tables in Materials Literature ' at the SouthNLP Symposium. You can find the poster here.
  • [Feb 2024] Presented 'Using Large Language Models to Mine Materials Literature' at Materials Research Data Alliance (MARDA) Annual Meeting. The presentation is available on YouTube.
  • [Dec 2023] Presented 'Retrieval of Synthesis Parameters of Polymer Nanocomposites Using LLMs' at AI4MAT Workshop at NeurIPS (New Orleans) and ML4Molecules Workshop.
  • [Oct 2023] Interviewed by the Materials Research Data Alliance (MARDA) on building the future of materials research.
  • [Nov 2023] Attended OpenAI Developers Conference in San Francisco.
  • [Aug 2022] Got full-tuition and stipend support from AI for Understanding and Designing Materials program.
  • [Aug 2022] Started PhD at Duke.
🌟 Life Outside Research When I'm not working on research, you can find me: - 🧗‍♀️ Bouldering at the climbing gym - 🏃‍♀️ Running on saturdays - Follow my progress on Strava! - 🧘‍♀️ Practicing hot yoga

Selected Publications

  1. LM4Sci at COLM 2025
    Information Extraction from Diverse Charts in Materials Science
    Circi, D., Bradley, M., Blouir, S., Wilthan, B., Anastasopoulos, A., Shehu, A., Catherine, L., Dhingra, B.
    LLM for Scientific Discovery Workshop at COLM 2025
    PDF

  2. AI4MAT at ICLR
    Challenges and Vision For Standardization of Biopolymer Datasets for Machine Learning
    Lalonde, J. N., Circi, D., Marrone, B. L., Zauscher, S., Catherine, L.
    AI for Accelerated Materials Design - ICLR 2026
    PDF

  3. NAACL
    Evaluating Morphological Compositional Generalization in Large Language Models
    Ismayilzada, M., Circi, D., Saleva J., Sirin, H., Koksal, A., Dhingra, B., Bosselut A., Plas, L, Ataman, D.
    Annual Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics, 2025
    PDF Code

  4. IMMI
    How Well Do Large Language Models Understand Tables in Materials Science?
    Circi, D., Khalighinejad, G., Chen, A., Dhingra, B., & Brinson, L.
    Integrating Materials and Manufacturing Innovation, 2024
    PDF Code

  5. AI4Materials at NeurIPS
    Retrieval of Synthesis Parameters of Polymer Nanocomposites Using LLMs
    Circi, D., Khalighinejad, G., Badhwar, S., Dhingra, B., & Brinson, L.
    AI4Materials Workshop at NeurIPS, 2023
    PDF

  6. Digital Discovery
    14 Examples of How LLMs Can Transform Materials Science and Chemistry
    Jablonka, K.M., Ai, Q., Al-Feghali, A., Badhwar, S., Bocarsly, J.D., Bran, A.M., Bringuier, S., Brinson, L.C., Choudhary, K., Circi, D., & Cox, S. et al.
    Digital Discovery, 2023
    PDF Code Won first place!

  7. Viruses
    The Mutation Profile of SARS-CoV-2 Is Primarily Shaped by the Host Antiviral Defense
    Azgari, C., Kilinc, Z., Turhan, B., Circi, D.., & Adebali, O.
    Viruses, 2021
    PDF Code

Services

Reviewing Roles

Organizer and Volunteer Roles