pith. sign in

arxiv: 2606.00915 · v1 · pith:775FC3CVnew · submitted 2026-05-30 · ⚛️ physics.optics

Autonomous agentic design for photonics

Prashanta Kharel , Amin Khavasi , Xinzhong Chen , Tyler W. Hughes This is my paper

Pith reviewed 2026-06-28 17:55 UTC · model grok-4.3

classification ⚛️ physics.optics
keywords photonic device designlarge language modelsautonomous agentssilicon photonicsnumerical simulationdevice optimizationmicroring modulatorsagentic design
0
0 comments X

The pith

Large language model agents autonomously design photonic devices to state-of-the-art performance by running propose-simulate-evaluate-iterate loops.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes that large language models instructed with access to numerical simulation tools and quantitative acceptance criteria can run autonomous design loops to create photonic devices. A sympathetic reader would care because the method shifts device creation from expert manual iteration to automated agent processes that handle passive components, active modulators, RF elements, and full chip layouts. Demonstrations show agents producing a complete silicon photonic modulator by combining layout, optical, charge, and electrode designs. The approach is presented as generalizing to any design task that pairs a simulator with criteria an LLM can judge.

Core claim

We introduce an automated, agent-driven approach to the design of photonic devices. We instruct large language models to solve photonic design problems, given access to software tools for performance evaluation through numerical simulations and quantitative acceptance criteria such as fabrication rules and physical-consistency checks. Within this context, agents run autonomous design loops of propose, simulate, evaluate, iterate and generate devices with state-of-the-art performance. We demonstrate this on passive components, active devices such as silicon microring modulators, RF devices such as traveling-wave electrodes, chip layout, and a combined silicon photonic modulator incorporating

What carries the argument

The autonomous agent design loop in which LLMs propose candidate devices, invoke simulation tools for performance evaluation, apply quantitative acceptance criteria, and iterate until targets are met.

If this is right

  • Agents produce state-of-the-art passive photonic components such as waveguide bends, splitters, and crossings.
  • Active devices including silicon microring modulators reach target performance through autonomous iteration.
  • RF devices such as traveling-wave electrodes for Mach-Zehnder modulators and electrical routing layouts are generated similarly.
  • A complete silicon photonic modulator can be assembled by one agent process covering layout, charge transport, optical mode, and RF electrode design.
  • The same loop applies to any design task that supplies a numerical simulator and criteria an LLM can evaluate.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could allow smaller teams to explore custom photonic designs that previously required large expert groups.
  • Similar agent loops might transfer to other simulator-driven fields such as electronics or mechanics if equivalent evaluation tools are supplied.
  • Performance gains would likely increase with improvements in the underlying simulators provided to the agents.
  • The combined modulator example suggests agents can coordinate multi-physics constraints in a single workflow.

Load-bearing premise

LLM agents can reliably interpret simulation outputs and fabrication rules to produce designs that meet or exceed human-designed performance without hidden post-hoc selection or excessive unreported compute.

What would settle it

Applying the agent system to a held-out photonic design problem and observing that no generated device meets the stated performance criteria after a comparable number of iterations would falsify the claim.

Figures

Figures reproduced from arXiv: 2606.00915 by Amin Khavasi, Prashanta Kharel, Tyler W. Hughes, Xinzhong Chen.

Figure 1
Figure 1. Figure 1: Agentic photonic design automation. (a) Photonic chip illustrating the design problems addressed here, with zoom-in views of each: electrical routing, MRMs, passive devices, MZMs, and RF electrodes. (b) The agentic photonic design automation loop, showing the design, verification, and simulation cycle that the agent enters to iterate toward a design goal set by humans under a set of constraints. 3 Results … view at source ↗
Figure 2
Figure 2. Figure 2: Passive bend. (a) Bend loss versus experiment over the 50-experiment budget (red crosses: discarded attempts; green circles: promoted to the running best; green curve: best-so-far envelope). (b) Wavelength dependence of the optimized bend loss, with the |E| field profile at 1550 nm (inset). White curve outlines the shape of the bend. 3.2 Active devices Next, we focus on the design of a silicon photonic mod… view at source ↗
Figure 3
Figure 3. Figure 3: Active devices. (a) 3-D view of the silicon microring modulator with a lateral PN phase shifter. (b) The agent’s 39 iterations in the (Cj , VπL) plane against the published silicon￾MRM compilation and its fit [30]; the agent traces the same envelope without being shown it, with iteration 28 (pink star) the best VπL · Cj . (c) Design doping and net free-carrier distribution at 1 V reverse bias for that desi… view at source ↗
Figure 4
Figure 4. Figure 4: RF electrode. (a) 3-D view of the segmented coplanar electrode. (b) Top-down and cross-section of one optimized iteration. (c) Run trajectory in the (Z0, α0) plane across the five topology families; the wide-cap T family (best design marked) clusters closest to the 50 Ω, low-loss, velocity-matched target. 3.4 Electrical routing External electrical signals need to be delivered to the various active componen… view at source ↗
Figure 5
Figure 5. Figure 5: Electrical routing. Starting layout (left): 192 DRC violations from default per-pin autorouting. Final layout (right): zero violations after 27 agentic iterations. 3.5 End-to-end multiphysics modulator In this section, we combine elements of all previous individual design problems into a single demon￾stration of a photonic device that couples all of these regimes. Here, we reproduce a published silicon pho… view at source ↗
Figure 6
Figure 6. Figure 6: End-to-end multiphysics modulator. (a) 3-D render of the T-rail traveling-wave silicon Mach–Zehnder modulator reproduced in this section. (b) Bandwidth versus modulation efficiency for the nine closed-loop designs; each is one fully optimized device, all holding Z0 within ±10% of 50 Ω. 4 Discussion and conclusion We have shown that a single agentic loop can coordinate components (passive, active, RF), layo… view at source ↗
read the original abstract

We introduce an automated, agent-driven approach to the design of photonic devices. We instruct large language models (LLMs) to solve photonic design problems, given access to software tools for performance evaluation (through numerical simulations) and quantitative acceptance criteria (e.g., fabrication rules, geometric constraints, physical-consistency checks). Within this context, agents run autonomous design loops (propose, simulate, evaluate, iterate) and generate devices with state-of-the-art performance. We demonstrate this approach in two stages: First, we run it individually on four canonical problem classes in photonic chip design: a) passive components (waveguide bends, splitters, crossings, etc.); b) active devices (silicon microring modulators (MRMs)); c) radio-frequency (RF) devices (traveling-wave electrodes for a Mach-Zehnder modulator (MZM)); d) chip layout (electrical routing). Then, we combine the previous studies in one demonstration to produce a silicon photonic modulator, incorporating layout, charge transport, optical mode, and RF electrode design. The approach generalizes to any problem that combines a numerical simulator with performance criteria that an LLM can evaluate.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces an agent-driven photonic design method in which LLMs are given access to numerical simulators and quantitative acceptance criteria (fabrication rules, geometric constraints, physical-consistency checks) and then run autonomous propose-simulate-evaluate-iterate loops. The approach is demonstrated first on four separate problem classes (passive components, silicon microring modulators, traveling-wave RF electrodes, and electrical routing) and then on a single integrated silicon photonic modulator that combines layout, charge transport, optical mode, and RF electrode design. The central claim is that the resulting devices achieve state-of-the-art performance and that the method generalizes to any simulator-plus-LLM-evaluable-criterion problem.

Significance. If the performance claims were quantitatively substantiated, the work would represent a notable step toward fully automated, multi-physics photonic design that could reduce reliance on expert-driven optimization loops. The integration of four distinct design domains into one modulator device is a non-trivial demonstration of the framework's scope. No machine-checked proofs, reproducible code releases, or parameter-free derivations are reported.

major comments (2)
  1. [Abstract] Abstract: the claim that agents 'generate devices with state-of-the-art performance' on four canonical classes plus an integrated modulator is unsupported by any quantitative metrics, error bars, baseline comparisons, success rates, iteration counts, or reporting of discarded runs. This absence is load-bearing for the central claim of autonomous superiority.
  2. [Abstract] Abstract / method description: the procedure is presented solely as an LLM-driven loop without equations, fitted parameters, or explicit description of how simulation outputs (mode profiles, S-parameters, charge transport, RF responses) are parsed and how fabrication constraints are enforced. This leaves the reliability of the LLM interpretation step unverified and prevents assessment of hidden post-hoc selection.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments. We address each major comment point by point below and indicate the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that agents 'generate devices with state-of-the-art performance' on four canonical classes plus an integrated modulator is unsupported by any quantitative metrics, error bars, baseline comparisons, success rates, iteration counts, or reporting of discarded runs. This absence is load-bearing for the central claim of autonomous superiority.

    Authors: We agree that the abstract's assertion of state-of-the-art performance requires quantitative substantiation that is not fully present in the current text. The revised manuscript will add specific performance metrics for each of the four problem classes and the integrated modulator, direct comparisons to published baselines or standard optimization approaches, error bars or statistical measures from repeated runs, success rates, iteration statistics, and a complete accounting of all runs including discarded designs. These additions will be placed in both the abstract and the results sections to support the central claim. revision: yes

  2. Referee: [Abstract] Abstract / method description: the procedure is presented solely as an LLM-driven loop without equations, fitted parameters, or explicit description of how simulation outputs (mode profiles, S-parameters, charge transport, RF responses) are parsed and how fabrication constraints are enforced. This leaves the reliability of the LLM interpretation step unverified and prevents assessment of hidden post-hoc selection.

    Authors: The manuscript currently describes the agentic loop at a conceptual level. We accept that greater detail is needed on output parsing and constraint enforcement. The revised version will include an expanded methods section specifying how simulation outputs are parsed into quantitative scores, how fabrication and physical-consistency constraints are encoded as evaluable criteria, the structure of the prompts and evaluation functions, and any post-processing steps. We will also document the full set of runs performed to allow assessment of selection effects. revision: yes

Circularity Check

0 steps flagged

No circularity: procedural demonstration without equations or self-referential reductions

full rationale

The paper presents an agent-driven procedural workflow for photonic device design using LLMs with simulation tools and acceptance criteria. No equations, fitted parameters, or mathematical derivations are described that could reduce to their own inputs by construction. The central claim is an empirical demonstration across device classes rather than a derivation chain. No self-citation load-bearing steps, uniqueness theorems, or ansatzes are invoked to justify core results. The method is self-contained as a description of autonomous loops; performance claims rest on reported outcomes rather than tautological redefinitions or fitted inputs renamed as predictions.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract supplies no explicit free parameters, axioms, or invented entities; the method is described at the level of a procedural framework rather than a mathematical model.

pith-pipeline@v0.9.1-grok · 5736 in / 1177 out tokens · 17953 ms · 2026-06-28T17:55:51.629508+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 29 canonical work pages · 2 internal anchors

  1. [1]

    G. T. Reed, G. Mashanovich, F. Y. Gardes, and D. J. Thomson. Silicon optical modulators. Nature Photonics, 4(8):518–526, 2010. doi: 10.1038/nphoton.2010.179. 20 Figure S4: Agent-built parametric layout of the Zhuang T-rail MZM.(a)Top view of the T-loaded CPS with two push–pull ribs.(b)Annotated 50µm unit cell.(c)Vertical stack and rib cross- section with ...

    work page doi:10.1038/nphoton.2010.179 2010

  2. [2]

    \ volume 12 ,\ pages 1700237 ( year 2018 ) NoStop

    Wim Bogaerts and Lukas Chrostowski. Silicon photonics circuit design: Methods, tools and challenges.Laser & Photonics Reviews, 12(4):1700237, 2018. doi: 10.1002/lpor.201700237

    work page doi:10.1002/lpor.201700237 2018

  3. [3]

    Torrijos-Mor´ an and D

    L. Torrijos-Mor´ an and D. P´ erez-L´ opez. Industry insight: photonics to scale AI data centers. npj Nanophotonics, 3:8, 2026. doi: 10.1038/s44310-025-00105-1

    work page doi:10.1038/s44310-025-00105-1 2026

  4. [4]

    Nanxi Li, C. P. Ho, J. Xue, L. W. Lim, G. Chen, Y. H. Fu, and L. Y. T. Lee. A progress review on solid-state LiDAR and nanophotonics-based LiDAR sensors.Laser & Photonics Reviews, 16:2100511, 2022. doi: 10.1002/lpor.202100511

    work page doi:10.1002/lpor.202100511 2022

  5. [5]

    Augmented reality and virtual reality displays: emerging technologies and future perspectives.Light: Science & Applications, 10:216, 2021

    Jianghao Xiong, En-Lin Hsiang, Ziqian He, Tao Zhan, and Shin-Tson Wu. Augmented reality and virtual reality displays: emerging technologies and future perspectives.Light: Science & Applications, 10:216, 2021. doi: 10.1038/s41377-021-00658-8

    work page doi:10.1038/s41377-021-00658-8 2021

  6. [6]

    Bowers, Alexis Bjorlin, Robert Blum, and John E

    Near Margalit, Chao Xiang, Steven M. Bowers, Alexis Bjorlin, Robert Blum, and John E. Bowers. Perspective on the future of silicon photonics and electronics.Applied Physics Letters, 118(22):220501, 2021. doi: 10.1063/5.0050117

    work page doi:10.1063/5.0050117 2021

  7. [7]

    Kazanskiy, Muhammad A

    Nikolay L. Kazanskiy, Muhammad A. Butt, and Svetlana N. Khonina. Optical computing: Status and perspectives.Nanomaterials, 12(13):2171, 2022. doi: 10.3390/nano12132171

    work page doi:10.3390/nano12132171 2022

  8. [8]

    Hughes, Momchil Minkov, Victor Liu, Zongfu Yu, and Shanhui Fan

    Tyler W. Hughes, Momchil Minkov, Victor Liu, Zongfu Yu, and Shanhui Fan. A perspective on the pathway toward full wave simulation of large area metalenses.Applied Physics Letters, 119(15):150502, 2021. doi: 10.1063/5.0071245. 25

    work page doi:10.1063/5.0071245 2021

  9. [9]

    Piggott, Weiliang Jin, Jelena Vuˇ ckovi´ c, and Alejandro W

    Sean Molesky, Zin Lin, Alexander Y. Piggott, Weiliang Jin, Jelena Vuˇ ckovi´ c, and Alejandro W. Rodriguez. Inverse design in nanophotonics.Nature Photonics, 12(11):659–670, 2018. doi: 10.1038/s41566-018-0246-9

    work page doi:10.1038/s41566-018-0246-9 2018

  10. [10]

    Hughes, Momchil Minkov, Ian A

    Tyler W. Hughes, Momchil Minkov, Ian A. D. Williamson, and Shanhui Fan. Adjoint method and inverse design for nonlinear nanophotonic devices.ACS Photonics, 5(12):4781–4787, 2018. doi: 10.1021/acsphotonics.8b01522

    work page doi:10.1021/acsphotonics.8b01522 2018

  11. [11]

    N. N. Shi, K. F. Chung, J. Tsai, Y. Augenstein, B. Zhang, T. W. Hughes, et al. Adjoint- optimized dual-layer grating couplers for low-loss, high-bandwidth optical interconnects. In Optical Fiber Communication Conference (OFC) 2026. Optica Publishing Group, 2026. Paper Tu2J.2

    2026

  12. [12]

    Robert Lupoiu, Yixuan Shao, Tianxiang Dai, Chenkai Mao, Kofi Edee, and Jonathan A. Fan. A multi-agentic framework for real-time, autonomous freeform metasurface design.Science Advances, 11(44):eadx8006, 2025. doi: 10.1126/sciadv.adx8006. arXiv:2503.20479

    work page doi:10.1126/sciadv.adx8006 2025

  13. [13]

    Yi Huang, Bowen Zheng, Yunxi Dong, Hong Tang, Huan Zhao, S. M. Rakibul Hasan Shawon, and Hualiang Zhang. A self-evolving agentic framework for metasurface inverse design.arXiv preprint arXiv:2604.01480, 2026

    work page arXiv 2026

  14. [14]

    Agentic metasurface design with self-correcting language-model systems

    Bei Wu, Bo Xiong, Haiyao Luo, Yaqi Li, Li Zhang, Qiaolu Chen, Hongsheng Chen, and Yihao Yang. Agentic metasurface design with self-correcting language-model systems.arXiv preprint arXiv:2605.22647, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  15. [15]

    Malof, and Willie J

    Darui Lu, Jordan M. Malof, and Willie J. Padilla. An agentic framework for autonomous metamaterial modeling and inverse design.ACS Photonics, 12(11):6071–6080, 2025. doi: 10.1021/acsphotonics.5c01514

    work page doi:10.1021/acsphotonics.5c01514 2025

  16. [16]

    Nanophotonic device design based on large language models: multilayer and metasurface examples.Nanophotonics, 14(8):1273–1282,

    Myungjoon Kim, Hyeonjin Park, and Jonghwa Shin. Nanophotonic device design based on large language models: multilayer and metasurface examples.Nanophotonics, 14(8):1273–1282,

  17. [17]

    doi: 10.1515/nanoph-2024-0674

    work page doi:10.1515/nanoph-2024-0674 2024

  18. [18]

    Multi-agent reinforcement learning for inverse design in photonic inte- grated circuits.arXiv preprint arXiv:2506.18627, 2025

    Yannik Mahlau, Maximilian Schier, Christoph Reinders, Frederik Schubert, Marco B¨ ugling, and Bodo Rosenhahn. Multi-agent reinforcement learning for inverse design in photonic inte- grated circuits.arXiv preprint arXiv:2506.18627, 2025

    work page arXiv 2025

  19. [19]

    Englund, and Joyce K

    Ankita Sharma, Yuqi Fu, Vahid Ansari, Rishabh Iyer, Fiona Kuang, Kashish Mistry, Raisa Is- lam Aishy, Sara Ahmad, Joaquin Matres, Dirk R. Englund, and Joyce K. S. Poon. AI agents for photonic integrated circuit design automation.APL Machine Learning, 3(4):046113, 2025. doi: 10.1063/5.0300741

    work page doi:10.1063/5.0300741 2025

  20. [20]

    PICBench: Bench- marking LLMs for photonic integrated circuits design

    Yuchao Wu, Xiaofei Yu, Hao Chen, Yang Luo, Yeyu Tong, and Yuzhe Ma. PICBench: Bench- marking LLMs for photonic integrated circuits design. InProceedings of the Design, Automa- tion and Test in Europe Conference (DATE), 2025. doi: 10.23919/DATE64628.2025.10992854

    work page doi:10.23919/date64628.2025.10992854 2025

  21. [21]

    Hongjian Zhou, Pingchuan Ma, and Jiaqi Gu. Toward intelligent electronic-photonic design automation for large-scale photonic integrated circuits: from device inverse design to physical layout generation.arXiv preprint arXiv:2507.22301, 2025

    work page arXiv 2025

  22. [22]

    Series push-pull mach–zehnder modulator with slow-wave t-rail electrodes: Reference implementation, 2026

    Amin Khavasi. Series push-pull mach–zehnder modulator with slow-wave t-rail electrodes: Reference implementation, 2026. This work, supplementary material. 26

    2026

  23. [23]

    autoresearch: AI agents running research on single-GPU nanochat training automatically.https://github.com/karpathy/autoresearch, 2026

    Andrej Karpathy. autoresearch: AI agents running research on single-GPU nanochat training automatically.https://github.com/karpathy/autoresearch, 2026

    2026

  24. [24]

    Song, C.A

    Daniil A. Boiko, Robert MacKnight, Ben Kline, and Gabe Gomes. Autonomous chemical research with large language models.Nature, 624(7992):570–578, 2023. doi: 10.1038/s41586- 023-06792-0

    work page doi:10.1038/s41586- 2023

  25. [25]

    N. J. Szymanski, B. Rendy, Y. Fei, R. E. Kumar, T. He, D. Milsted, M. J. McDermott, M. Gallant, E. D. Cubuk, A. Merchant, H. Kim, A. Jain, C. J. Bartel, K. Persson, Y. Zeng, and G. Ceder. An autonomous laboratory for the accelerated synthesis of inorganic materials. Nature, 624(7990):86–91, 2023. doi: 10.1038/s41586-023-06734-w

    work page doi:10.1038/s41586-023-06734-w 2023

  26. [26]

    Pawan Kumar, Emilien Dupont, Francisco J

    Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M. Pawan Kumar, Emilien Dupont, Francisco J. R. Ruiz, Jordan S. Ellenberg, Pengming Wang, Omar Fawzi, Pushmeet Kohli, and Alhussein Fawzi. Mathematical discoveries from program search with large language models.Nature, 625(7995):468–475, 2024. doi: 10.1038/ s41586-023-06924-6

    2024

  27. [27]

    Schubert, Alfred K

    Martin F. Schubert, Alfred K. C. Cheung, Ian A. D. Williamson, Aleksandra Spyra, and David H. Alexander. Inverse design of photonic devices with strict foundry fabrication con- straints.ACS Photonics, 9(7):2327–2336, 2022. doi: 10.1021/acsphotonics.2c00313

    work page doi:10.1021/acsphotonics.2c00313 2022

  28. [28]

    Cherchi, S

    M. Cherchi, S. Ylinen, M. Harjanne, M. Kapulainen, and T. Aalto. Dramatic size reduction of waveguide bends on a micron-scale silicon photonic platform.Optics Express, 21(15):17814– 17823, 2013. doi: 10.1364/OE.21.017814

    work page doi:10.1364/oe.21.017814 2013

  29. [29]

    Analysis of silicon nitride partial Euler waveguide bends.Optics Express, 27(22):31394–31406, 2019

    Florian Vogelbacher, Stefan Nevlacsil, Martin Sagmeister, Jochen Kraft, Karl Unterrainer, and Rainer Hainberger. Analysis of silicon nitride partial Euler waveguide bends.Optics Express, 27(22):31394–31406, 2019. doi: 10.1364/OE.27.031394

    work page doi:10.1364/oe.27.031394 2019

  30. [30]

    Universal design of waveguide bends in silicon-on-insulator photonics platform.Journal of Lightwave Technology, 37(13):3044–3054, 2019

    Meisam Bahadori, Mahdi Nikdast, Qixiang Cheng, and Keren Bergman. Universal design of waveguide bends in silicon-on-insulator photonics platform.Journal of Lightwave Technology, 37(13):3044–3054, 2019. doi: 10.1109/JLT.2019.2909983

    work page doi:10.1109/jlt.2019.2909983 2019

  31. [31]

    Si microring resonator modulators at>200 Gb/s

    David Patel. Si microring resonator modulators at>200 Gb/s. InOptical Fiber Communica- tion Conference (OFC) 2026. Optica Publishing Group, 2026. Paper M2A.7

    2026

  32. [32]

    Breaking voltage–bandwidth limits in integrated lithium niobate modulators using micro-structured electrodes.Optica, 8(3):357–363, 2021

    Prashanta Kharel, Christian Reimer, Kevin Luke, Lingyan He, and Mian Zhang. Breaking voltage–bandwidth limits in integrated lithium niobate modulators using micro-structured electrodes.Optica, 8(3):357–363, 2021. doi: 10.1364/OPTICA.416155

    work page doi:10.1364/optica.416155 2021

  33. [33]

    Equivalent circuit model of the carrier- depletion-based push–pull silicon optical modulators with T-rail slow wave electrodes.IEEE Photonics Journal, 16(4):5500809, 2024

    Dongwei Zhuang, Quanxin Na, Qijie Xie, Nan Zhang, Lanxuan Zhang, Xin Li, Guomeng Zuo, Hao Zhang, Lei Wang, Li Qin, and Junfeng Song. Equivalent circuit model of the carrier- depletion-based push–pull silicon optical modulators with T-rail slow wave electrodes.IEEE Photonics Journal, 16(4):5500809, 2024. doi: 10.1109/JPHOT.2024.3427830

    work page doi:10.1109/jphot.2024.3427830 2024

  34. [34]

    FermiLink: A Unified Agent Framework for Multidomain Autonomous Scientific Simulations

    Gang Meng, Andres Felipe Bocanegra Vargas, Xinwei Ji, Federico Garcia-Gaitan, Felipe Reyes- Osorio, Jalil Varela-Manjarres, Yafei Ren, Mohammadhasan Dinpajooh, Branislav K. Nikoli´ c, and Tao E. Li. FermiLink: A unified agent framework for multidomain autonomous scientific simulations.arXiv preprint arXiv:2604.03460, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  35. [35]

    Soref and Brian R

    Richard A. Soref and Brian R. Bennett. Electrooptical effects in silicon.IEEE Journal of Quantum Electronics, 23(1):123–129, 1987. doi: 10.1109/JQE.1987.1073206. 27

    work page doi:10.1109/jqe.1987.1073206 1987

  36. [36]

    Design, analysis, and performance of a silicon photonic traveling wave Mach– Zehnder modulator

    David Patel. Design, analysis, and performance of a silicon photonic traveling wave Mach– Zehnder modulator. Master of engineering thesis, McGill University, 2015. URLhttps:// escholarship.mcgill.ca/concern/theses/gm80hz52p

    2015

  37. [37]

    Agentic photonic design: Designing ten modulators overnight with a multi- physics agent in the loop

    Amin Khavasi. Agentic photonic design: Designing ten modulators overnight with a multi- physics agent in the loop. Flexcompute Engineering Blog,https://hs.flexcompute.com/ blog/designing-modulators; code, plan, operating rules, and append-only journals athttps: //github.com/aminkhavasi/modulator-autodesign, 2026. Published 2026-05-13. 28

    2026