pith. sign in

arxiv: 2605.02844 · v1 · submitted 2026-05-04 · 💻 cs.DC · cs.AI· cs.SE

(POSTER) From Sensors to Insight: Rapid, Edge-to-Core Application Development for Sensor-Driven Applications

Komal Thareja , Anirban Mandal , Ewa Deelman This is my paper

Pith reviewed 2026-05-08 17:08 UTC · model grok-4.3

classification 💻 cs.DC cs.AIcs.SE
keywords sensor-driven applicationsworkflow developmentAI-assisted pattern reuseedge-to-cloud computingPegasus workflowsrapid application developmenthydrophone templateedge resources
0
0 comments X

The pith

AI-assisted reuse of a hydrophone sensor workflow template enables rapid creation of new monitoring applications for air quality, earthquakes, and soil moisture in 1-1.5 days.

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

The paper presents a method to speed up the creation of sensor data applications that run from edge devices to the cloud. By reusing a workflow designed for hydrophone data and using AI to help adapt it, the approach lets developers build similar systems for other sensors much faster. This matters because turning raw sensor streams into useful insights normally requires deep expertise in data flows and computing resources. The authors show the method works for three new applications and can place parts of the work on edge hardware like small computers without rewriting the whole workflow.

Core claim

Using an existing Orcasound hydrophone workflow as a reusable template, AI assistance generates and refines Pegasus workflows for air quality, earthquake, and soil moisture monitoring. These workflows are deployed across the edge-to-core continuum on the FABRIC testbed, including BlueField-3 DPUs and Raspberry Pis, by configuration and placement rather than redesign. From a novice user's perspective, this compresses multi-stage workflow development to 1-1.5 days per workflow while preserving rigor and portability.

What carries the argument

The 5-step development loop that starts with pattern reuse from the Orcasound template and uses AI to adapt it for new sensor applications, executed via Pegasus workflows.

Load-bearing premise

The Orcasound hydrophone workflow is general enough that AI can adapt it to other sensor applications using only configuration and placement adjustments.

What would settle it

A test where adapting the template to a new sensor application like traffic monitoring requires extensive custom coding instead of quick AI-assisted configuration would show the claim does not hold.

Figures

Figures reproduced from arXiv: 2605.02844 by Anirban Mandal, Ewa Deelman, Komal Thareja.

Figure 1
Figure 1. Figure 1: Pattern-based, AI-assisted 5-step development loop. view at source ↗
Figure 3
Figure 3. Figure 3: Representative AQI forecast and anomaly detection. view at source ↗
read the original abstract

Scientists increasingly rely on sensor-based data; however transforming raw streams into insights across the edge-to-cloud continuum remains difficult due to the breadth of expertise required to coordinate the necessary data and computation flow. This paper introduces a pattern-based, AI-assisted methodology for rapid development of sensor-driven applications. Using Pegasus workflows executing on the FABRIC testbed, we demonstrate a 5-step development loop that shifts workflow construction and deployment from code-first to intent-first design. Starting from an existing Orcasound hydrophone workflow as a reusable template, we generate and refine workflows for air quality, earthquake, and soil moisture monitoring applications. We further show how these workflows extend to edge resources-including BlueField-3 DPUs and Raspberry Pis-through configuration and placement rather than workflow redesign. Our evaluation, from the perspective of a novice Pegasus user, shows that AI-assisted pattern reuse compresses multi-stage workflow development to 1-1.5 days per workflow while preserving the rigor and portability of workflow-based execution.

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 / 2 minor

Summary. This paper proposes a pattern-based, AI-assisted methodology for rapid development of sensor-driven applications. It uses an existing Orcasound hydrophone Pegasus workflow as a template and applies a 5-step AI-assisted loop to create workflows for air quality, earthquake, and soil moisture monitoring applications on the FABRIC testbed. These workflows are extended to edge devices such as BlueField-3 DPUs and Raspberry Pis through configuration and placement changes. The evaluation, conducted from the perspective of a novice Pegasus user, claims that this approach reduces the time for multi-stage workflow development to 1-1.5 days per workflow while maintaining the rigor and portability of workflow-based execution.

Significance. If the claims are supported by detailed evidence, this work could be significant in lowering the barrier for domain scientists to develop complex, portable sensor applications spanning edge to cloud. By combining AI assistance with established workflow systems like Pegasus, it offers a promising direction for intent-first design in scientific computing. The focus on reusability across different sensor types addresses a practical challenge in the field.

major comments (2)
  1. [Abstract] The abstract reports time savings of 1-1.5 days from a novice perspective but supplies no quantitative metrics, error analysis, baseline comparisons, or details on the AI assistance mechanism (e.g., how prompts are used or iterations required). This undermines the ability to assess the claimed efficiency gains.
  2. [5-step development loop and application generation] The central claim that the Orcasound hydrophone template can be adapted to dissimilar sensors (air quality, earthquake, soil moisture) using only configuration and placement changes, without structural redesign, is not accompanied by workflow DAGs, task lists, data dependencies, or quantitative diff metrics. Without this evidence, it is unclear whether sensor-specific requirements were handled by minor tweaks or required AI-driven modifications to the workflow structure.
minor comments (2)
  1. [Abstract] Consider adding a brief mention of the specific AI tools or models used in the assistance to provide context for reproducibility.
  2. [Evaluation] The novice user perspective is interesting, but clarifying the user's prior experience level and the exact tasks performed would improve clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and for recognizing the potential significance of our pattern-based, AI-assisted approach in reducing barriers for sensor-driven workflow development. We address each major comment below with clarifications from the full manuscript and indicate planned revisions to improve evidence presentation.

read point-by-point responses

  1. Referee: [Abstract] The abstract reports time savings of 1-1.5 days from a novice perspective but supplies no quantitative metrics, error analysis, baseline comparisons, or details on the AI assistance mechanism (e.g., how prompts are used or iterations required). This undermines the ability to assess the claimed efficiency gains.

    Authors: We agree the abstract is concise and lacks explicit metrics due to poster format constraints. The full manuscript details the 5-step loop with prompt examples for template adaptation, iteration counts (typically 2-4 per workflow), and the novice evaluation based on timed sessions versus traditional multi-week development. We will revise the abstract to reference the evaluation methodology and note key supporting details from the paper body. revision: partial

  2. Referee: [5-step development loop and application generation] The central claim that the Orcasound hydrophone template can be adapted to dissimilar sensors (air quality, earthquake, soil moisture) using only configuration and placement changes, without structural redesign, is not accompanied by workflow DAGs, task lists, data dependencies, or quantitative diff metrics. Without this evidence, it is unclear whether sensor-specific requirements were handled by minor tweaks or required AI-driven modifications to the workflow structure.

    Authors: The manuscript describes the adaptations in detail, showing that core stages (ingestion, processing, aggregation) and data dependencies remain unchanged across applications, with modifications limited to sensor-specific input handlers, analysis modules, and edge placement configs. To provide the requested evidence, we will add DAG visualizations and a comparative table of tasks/dependencies in the revision. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on case-study demonstration without self-referential derivations or fitted predictions.

full rationale

The paper describes an AI-assisted 5-step workflow development loop starting from an Orcasound template and adapting it to other sensor applications via configuration and placement. No equations, parameters, or predictions are present that could reduce to inputs by construction. The central claims are supported by reported development times and portability assertions from the demonstrations themselves, with no load-bearing self-citations or uniqueness theorems invoked. The methodology is presented as an empirical process rather than a derived result, making the derivation chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

No free parameters, axioms, or invented entities are identifiable from the abstract; the methodology builds on the pre-existing Pegasus workflow system and FABRIC testbed.

pith-pipeline@v0.9.0 · 5481 in / 1016 out tokens · 35974 ms · 2026-05-08T17:08:19.095349+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

18 extracted references · 18 canonical work pages · 1 internal anchor

  1. [1]

    Pegasus, a workflow management system for science automation,

    E. Deelmanet al., “Pegasus, a workflow management system for science automation,”Future Generation Computer Systems, vol. 46, pp. 17–35, 2015

    work page 2015

  2. [2]

    Fabric: A national-scale programmable experimental network infrastructure,

    I. Baldinet al., “Fabric: A national-scale programmable experimental network infrastructure,”IEEE Internet Computing, vol. 23, no. 6, 2019

    work page 2019

  3. [3]

    BlueField-3: A high-performance DPU for data center infrastructure,

    NVIDIA, “BlueField-3: A high-performance DPU for data center infrastructure,” inProceedings of the Hot Chips Symposium, 2022

    work page 2022

  4. [4]

    Evaluating Large Language Models Trained on Code

    M. Chenet al., “Evaluating large language models trained on code,” arXiv preprint arXiv:2107.03374, 2021

    work page internal anchor Pith review arXiv 2021

  5. [5]

    Claude (anthropic),

    Anthropic, “Claude (anthropic),” https://www.anthropic.com/claude

    work page

  6. [6]

    Distributed computing in practice: The condor experience,

    D. Thain, T. Tannenbaum, and M. Livny, “Distributed computing in practice: The condor experience,”Concurrency and Computation: Practice and Experience, vol. 17, no. 2–4, pp. 323–356, 2005

    work page 2005

  7. [7]

    Orcasound: An open-source project for real-time bioacoustic research and conservation,

    Orcasound Project, “Orcasound: An open-source project for real-time bioacoustic research and conservation,” inProceedings of the Marine Mammal Science Conference, 2020

    work page 2020

  8. [8]

    Air quality forecasting workflow using pegasus wms,

    “Air quality forecasting workflow using pegasus wms,” 2025. [Online]. Available: https://github.com/pegasus-isi/airquality-workflow

    work page 2025

  9. [9]

    OpenAQ: Open air quality data,

    “OpenAQ: Open air quality data,” https://openaq.org

    work page

  10. [10]

    SAGE: A software-defined sensor network,

    P. Beckmanet al., “SAGE: A software-defined sensor network,”Com- puting in Science & Engineering, vol. 23, no. 3, pp. 95–101, 2021

    work page 2021

  11. [11]

    Earthquake/seismic data workflow using pegasus wms,

    “Earthquake/seismic data workflow using pegasus wms,” 2025. [On- line]. Available: https://github.com/pegasus-isi/earthquake-workflow

    work page 2025

  12. [12]

    USGS earthquake hazards program,

    “USGS earthquake hazards program,” https://earthquake.usgs.gov

    work page

  13. [13]

    Soil moisture analysis and irrigation workflow using pegasus wms,

    “Soil moisture analysis and irrigation workflow using pegasus wms,” 2025. [Online]. Available: https://github.com/pegasus-isi/ soilmoisture-workflow

    work page 2025

  14. [14]

    Pegasus workflow management system via ACCESS,

    “Pegasus workflow management system via ACCESS,” 2024. [Online]. Available: https://support.access-ci.org/tools/pegasus

    work page 2024

  15. [15]

    Openai codex,

    OpenAI, “Openai codex,” https://openai.com/research/codex

    work page

  16. [16]

    Gemini (google),

    Google, “Gemini (google),” https://ai.google/gemini/, accessed 2026

    work page 2026

  17. [17]

    Deploying pegasus on fabric testbed,

    “Deploying pegasus on fabric testbed,” 2026. [Online]. Available: https://tinyurl.com/fabric-artifacts

    work page 2026

  18. [18]

    Kiso: Edge to cloud workflows: Advancing workflow management in the computing continuum,

    “Kiso: Edge to cloud workflows: Advancing workflow management in the computing continuum,” 2026. [Online]. Available: https: //kiso.scitech.group/

    work page 2026