Mehmet Yigit Turali

Engineering IV Building

UCLA

Los Angeles, CA 90095

I’m a M.S./Ph.D. student at the Department of Electrical and Computer Engineering, University of California, Los Angeles, advised by Prof. Vwani P. Roychowdhury. My research sits at the intersection of large language models, structured memory, and machine learning for healthcare.

Currently, I’m building LLM agents with structured memory, centered on Generative Semantic Workspaces (GSW) — entity- and event-aware memories of atomic question–answer pairs that models can reason over. I’m leading PANINI-DR (first author, under review at ICLR 2027), a multi-agent deep-research harness that separates coordinating an investigation from researching and verifying its subquestions, with a shared GSW tracking subquestions, dependencies, findings, and evidence. It reaches 73.3% accuracy on BrowseComp-Plus, 16 points above the next-best baseline, and retains the most recoverable evidence for its answers across BrowseComp-Plus, FRAMES, and MoNaCo. PANINI-DR builds on PANINI (ICML 2026), a non-parametric continual-learning framework that delivers competitive multi-hop QA accuracy with 2–30× fewer answer-context tokens than baselines.

In Summer 2026, I served as the Academic Mentor for the OpenAI team at the IPAM RIPS program at UCLA — partnering with OpenAI to advise a team of top undergraduates on an OpenAI-sponsored summer research project.

I also work on:

NLP for genomics: reference-free DNA embeddings and one-shot structural-variant detection treating DNA as a specialized language.
Brain-organoid computing: reservoir-computing approaches that harness biological neural networks for chaotic time-series forecasting, and time-resolved network analysis that distinguishes pathological states in organoid models (ICASSP 2026).
GlobalHealthLLM: LLM-driven standardization of pediatric epilepsy records from rural Ugandan clinics, supporting medication-decision tools — including MANANA (NeurIPS 2026), which learns local anti-seizure-medication prescribing guidance and defers uncertain cases to specialists.

Before UCLA, I earned my B.S. in Electrical and Electronics Engineering at Bilkent University (Summa Cum Laude, 2024), where I was a researcher under Prof. S. Serdar Kozat publishing on adaptive feature selection, gradient-boosted sequential regression, and SGD-based filtering. I’m a recipient of the Bilkent Research Excellence Award (2024) and the Responsible AI Innovation Award (MathWorks / FIGES / Tubitak Sage, 2024).

Feel free to reach out if you’d like to chat about LLM memory systems, deep-research agents, ML for health, or computational neuroscience.

news

Oct 10, 2026	📝 Serving as a reviewer for the NeurIPS 2026 Workshop on GenAI for Health (GenAI4Health): Agentic Systems, Clinical Trust, and Future Potential.
Oct 08, 2026	📚 TA’ing ECE C147/C247A: Neural Networks and Deep Learning at UCLA again this Fall 2026 quarter.
Oct 07, 2026	🎉 Our paper Teaching LLMs to Recommend and Defer in Underrepresented Epilepsy Care (MANANA) has been accepted to NeurIPS 2026! MANANA learns local anti-seizure-medication prescribing guidance for pediatric epilepsy care in Uganda and defers uncertain cases to specialists. Joint work led by S. Rajesh and K. Sharma with Prof. V. Roychowdhury, R. Mazumder, and collaborators. arXiv · code
May 31, 2026	🎓 Passed my Ph.D. Preliminary Exam at UCLA — officially advancing in the Ph.D. program!
May 04, 2026	🤝 Thrilled to share that I’ll be the Academic Mentor for the OpenAI team this summer at the IPAM RIPS program at UCLA (June 22 – August 21, 2026) — partnering with OpenAI to mentor a team of top undergraduates on an OpenAI-sponsored research project.

selected publications

Under Review
PANINI-DR: Controlled Deep Research Through Generative Semantic Workspaces

M. Y. Turali, S. Rajesh, M. E. Lorasdagi, and 3 more authors

In Submitted to the Fifteenth International Conference on Learning Representations (ICLR 2027), 2026

Under review.

Abs Bib HTML

Deep research requires connecting multiple steps of search, reasoning, and computation, maintaining research state across model calls, and checking findings before errors propagate. Panini-DR is a multi-agent harness that separates coordination of the overall investigation from the research and verification of individual subquestions. A shared workspace, built on the Generative Semantic Workspace representation, records subquestions, their dependencies, findings, and supporting evidence, while a separate verification agent checks proposed answers against that evidence. Evaluated on BrowseComp-Plus, FRAMES, and MoNaCo against seven baselines with a shared language-model backbone and frozen corpora, Panini-DR achieves the strongest average performance: 73.3% accuracy on BrowseComp-Plus, 16.0 points above the next-best baseline at higher inference cost, and 26 of 50 hard questions answered correctly, nearly twice the next-best baseline. It also achieves the highest answer recoverability from saved evidence on all three benchmarks.
@inproceedings{turali2026paninidr, title = {{PANINI-DR}: Controlled Deep Research Through Generative Semantic Workspaces}, author = {Turali, M. Y. and Rajesh, S. and Lorasdagi, M. E. and Duan, C. and Sharma, K. and Roychowdhury, V.}, booktitle = {Submitted to the Fifteenth International Conference on Learning Representations (ICLR 2027)}, year = {2026}, note = {Under review.}, }
NeurIPS
Teaching LLMs to Recommend and Defer in Underrepresented Epilepsy Care

S. Rajesh, K. Sharma, T. Monsoor, and 8 more authors

In Advances in Neural Information Processing Systems (NeurIPS), 2026

*Equal contribution by Rajesh and Sharma.

Abs arXiv Bib Code

LLM-based decision support for anti-seizure medication management in Ugandan pediatric epilepsy care must adapt to local prescribing practice and know when to defer. MANANA is a non-parametric prompt-learning framework, with single- and multi-agent variants, that converts observed prescription errors into auditable prompt memories learned from a small patient-level training set. Bayesian prompt averaging turns the learned prompt trajectory into prescription likelihoods and an uncertainty-based deferral signal: on an independently collected held-out cohort, it improves visit-level top-3 prescription accuracy by 4-8 percentage points over prompt-optimization baselines and auto-handles the most confident half of cases at 95% precision while deferring the rest for specialist review.
@inproceedings{rajesh2026manana, title = {Teaching LLMs to Recommend and Defer in Underrepresented Epilepsy Care}, author = {Rajesh, S. and Sharma, K. and Monsoor, T. and Turali, M. Y. and Idro, R. and Kayaga, J. and Sebunya, R. and Namata, T. T. and Pasqua, J. N. and Roychowdhury, V. and Mazumder, R.}, booktitle = {Advances in Neural Information Processing Systems (NeurIPS)}, year = {2026}, note = {*Equal contribution by Rajesh and Sharma.}, }
ICML
Panini: Continual Learning in Token Space via Structured Memory

S. Rajesh, P. S. Holur, M. Y. Turali, and 2 more authors

In Proceedings of the 43rd International Conference on Machine Learning (ICML), 2026

Accepted as poster. *Equal contribution by Rajesh and Holur.

Abs Bib

PANINI is a non-parametric continual-learning framework that encodes documents into Generative Semantic Workspaces (GSW)—entity- and event-aware networks of atomic question-answer pairs. A beam-search retrieval procedure (Reasoning Inference Chain Retrieval, RICR) follows multi-hop inference chains through GSWs to assemble compact evidence, achieving top average performance across six QA benchmarks while using 2-30x fewer answer-context tokens than competitive baselines.
@inproceedings{rajesh2026panini, title = {Panini: Continual Learning in Token Space via Structured Memory}, author = {Rajesh, S. and Holur, P. S. and Turali, M. Y. and Duan, C. and Roychowdhury, V.}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning (ICML)}, year = {2026}, note = {Accepted as poster. *Equal contribution by Rajesh and Holur.} }
NeurIPS-W
Demo: Customizing Open-Source LLMs for Quantitative Medication Attribute Extraction across Heterogeneous EHR Systems

Z. Fei, M. Y. Turali, S. Rajesh, and 7 more authors

In NeurIPS 2025 Workshop on GenAI for Health: Potential, Trust, and Policy Compliance, 2025

Accepted as poster. *Equal contribution by Fei, Turali, and Rajesh.

Abs Bib

We customize open-source LLMs to extract quantitative medication attributes (dose, frequency, duration) from heterogeneous electronic health record systems, demonstrating robustness across formats relevant to addiction-medicine deployments.
@inproceedings{fei2025ehrllm, title = {Demo: Customizing Open-Source LLMs for Quantitative Medication Attribute Extraction across Heterogeneous EHR Systems}, author = {Fei, Z. and Turali, M. Y. and Rajesh, S. and Dai, X. and Pham, H. and Holur, P. S. and Zhu, Y. and Mooney, L. J. and Hser, Y.-I. and Roychowdhury, V.}, booktitle = {NeurIPS 2025 Workshop on GenAI for Health: Potential, Trust, and Policy Compliance}, year = {2025}, note = {Accepted as poster. *Equal contribution by Fei, Turali, and Rajesh.} }