AI Developments That Changed Vibrational Spectroscopy in 2025

Introduction

Artificial intelligence (AI) is promising to redefine vibrational spectroscopy in 2025, marking a major inflection point in spectroscopic analysis, calibration, and interpretation across Raman, infrared (IR), near-infrared (NIR), and hyperspectral imaging platforms. From agricultural sensing to precision oncology, the fusion of machine learning (ML), deep neural networks (NNs), quantile regression forests, and explainable AI (XAI) is transforming spectroscopic workflows into autonomous, scalable, and predictive modeling systems (1–3). New software platforms such as SpectrumLab and SpectraML demonstrate how generative models, foundation architectures, and physics-informed neural networks can automate feature extraction and deliver predictive models with actionable uncertainty estimates (1–2,4). The following review consolidates trends across 30 Spectroscopy publications from 2025, offering a narrative on how AI is enabling innovation in data fusion, spectral imaging, industrial bioprocess monitoring, precision agriculture, environmental risk assessment, biomedical diagnostics, and other applications.

Vibrational spectroscopy, rooted historically in classic multivariate chemometrics, now stands at the convergence of automation, robotics, and data-driven intelligence made possible by advancing computational power. Until recently, spectral interpretation required expert knowledge, careful pre-processing, and assumptions on sample structure. Today, AI-based models are capable of ingesting raw hyperspectral data cubes, extracting latent molecular signatures, and returning calibrated results within milliseconds (2,5,6). This transformation arises from the synergy between well-proven legacy chemometrics—principal components analysis (PCA), partial least squares (PLS), and multiple linear regression (MLR)—and modern AI, for example, convolutional neural networks (CNNs), transformers, quantile regression forests, and multimodal fusion networks (2,4,7). Research across 2025 reflects a rapid shift: spectroscopic instruments are becoming more than passive recorders—they are promising to become more intelligent, predictive, and self-optimizing systems (1–3,8).

Foundational Evolution: From Classical Chemometrics to Intelligent Models

Chemometrics and AI are no longer parallel frameworks—they are becoming a merged discipline (2,6). Early spectroscopy relied on PLS regression for quantitative analysis, and PCR/PCA for dimensionality reduction, data interpretation, and qualitative assessments. These and other techniques were used for curve-fitting for absorption peak interpretation (2,6). In 2025, calibration increasingly employs ensemble learners and deep neural networks capable of identifying nonlinear interactions, overlapping components, compositional intercorrelations, and spectral anomalies automatically (6,9,10). The historical development and projected future of calibration are captured in a series of two-part chemometrics reviews (1–2,6,22), showing how calibration now extends toward autonomous correction, real-time learning, and cross-instrument transfer using foundation models.

Explainable Artificial Intelligence (XAI) for Model Transparency

Spectroscopy’s longstanding challenge has not necessarily been accuracy—it has been interpretability. Black-box neural predictors risk misclassification in clinical or regulatory environments, making XAI essential (5). One article (5) outlines techniques such as SHapley Additive exPlanations (SHAP), and Gradient-weighted Class Activation Mapping (Grad-CAM) saliency maps, decision-boundary visualization, and latent-feature attribution, all having the potential to enable translation of hidden spectral dependencies into chemically meaningful markers. As spectroscopy enters medicine, environmental policy, and industrial compliance, transparency and interpretability frameworks will determine adoption as much as raw predictive accuracy performance (5,11).

Uncertainty-Aware Calibration: Quantile Regression Forests for Predictive Confidence

Wadoux and Ramirez-Lopez demonstrate a milestone: ML models that quantify their own uncertainty when predicting agricultural soil properties (4). Quantile regression forest (QRF) models estimate both a mean prediction and its statistical confidence, addressing a foundational industry need—knowing how much to trust the prediction (4). When applied to diffuse-reflectance soil spectroscopy, QRF enables risk-aware decision systems for resource allocation, irrigation planning, and environmental compliance (4,12–14).

AI-Accelerated Environmental Spectroscopy

Raman spectroscopy and AI are used for microplastics detection. Micro- and nanoplastics persist through global waterways, food systems, and waste streams (7,15).

Luo and coworkers (15) combine Raman spectral imaging with CNN-based classification, producing fast and highly selective polymer fingerprinting. A paired tutorial (7) details workflows, feature extraction, peak-library approaches, and automated classifier deployment for environmental lab or field studies. Together, they represent a major advance in pollutant surveillance and ecological risk modeling (7,15).

Raman Fingerprinting for Pesticides

A mini-tutorial describes pesticide classification in via a 785-nm Raman instrument and random forest modeling, achieving robust, field-deployable performance (16). The method supports agricultural screening, residue quantification, and regulatory compliance across multianalyte systems (14,16).

Satellite-Spectroscopy Fusion for River Pollution

Large-scale water quality monitoring previously required on-site chemical sampling and field or laboratory measurements; now, space-based satellites spectrometers using ML are able to infer pollution metrics such as chemical oxygen demand (COD), total phosphorus (TP), ammonia-nitrogen (NH₃-N), and dissolved oxygen in Chinese river basins (13). This represents a shift toward geospatial vibrational sensing integrated with cloud-scale analytics.

Agriculture and Soil Chemistry: Precision Spectroscopy at Field Scale

Drone hyperspectral sensing with ML enables sub-surface moisture estimation at 10–30 cm depth, improving irrigation, drought mitigation, and yield forecasting (12,14). Mid infrared (MIR) spectroscopy combined with CNN modeling quantifies particulate vs. mineral-bound soil organic carbon, accelerating climate and carbon-cycle research (10). These approaches support sustainable resource management and align with global emissions tracking initiatives (10,12–14).

Biomedical Spectroscopy and AI Diagnostics

Surface-enhanced Raman spectroscopy (SERS) platforms augmented with neural networks are reshaping medical diagnostics for biomedical analyte detection (8,17,18). A major review from Shanghai Jiao Tong University (8) shows how AI improves analyte sensitivity, enhances spectral denoising, decodes biochemical heterogeneity, and advances point-of-care diagnostics. Wearable SERS “smart skin” sensors detect biomarkers, hydration states, and metabolic fluctuations continuously—ushering a potential era of molecular-level wearable medicine (17).

Cancer Immunotherapy Guided by Raman and AI

Precision oncology requires real-time tumor phenotyping. AI-interpreted Raman spectra are now promising to map tumor microenvironments, supporting treatment selection and patient-specific immunotherapy strategies (11). This advancement represents one of 2025’s strongest clinical advances, where molecular vibrational signatures are used for therapeutic decision-making intelligence (8,11,17).

Medicinal/Edible Homolog Authentication

A surface-enhanced Raman spectroscopy (SERS) combined with an AI software platform (SERSome) integrates AI-classified Raman spectra for verification of medicinal and edible homologs, reducing counterfeiting and toxic substitutions (18). This intersection of food safety and biomedical spectroscopy sets a precedent for potential AI-regulated natural product authentication (18).

Industrial Bioprocessing and Autonomous Control Using Spectroscopy

In one application, dual NIR-Raman monitoring instrumentation optimized gentamicin fermentation conditions with real-time AI feedback (19). AI model-guided spectroscopy was shown to shorten process manufacturing cycles, reduce waste, and accurately scale industrial throughput. Similar frameworks now show promise in metabolite tracking, pharmaceutical purification, and reactor-closed-loop control processes (19,20).

NIR hyperspectral imaging is being tested for stand-off explosive and hazardous chemical identification, demonstrating powerful security applications with CNN classification accuracy in complex outdoor environments (20).

AI-Accelerated Fundamental Vibrational Physics

Two published Digital Discovery reviews (9,21) summarize how ML replaces costly quantum modeling in vibrational simulation. Neural surrogates predict phonon dispersions, anharmonic potentials, and molecular vibrational modes at a fraction of density functional theory/molecular dynamics (DFT/MD) computational cost, showing potential to revolutionize materials science research and development (9,21).

Future Directions: Foundational Models, Physics-Informed Learning & Autonomous Interpretation

This year marked a pivotal shift in computational spectroscopy, moving beyond narrowly trained models toward large, general-purpose AI systems that can analyze unknown spectra, generate new spectral patterns, and detect unusual samples without explicit prior training (1–3). These next-generation architectures function more like scientific partners than tools: they reason across modalities, infer chemistry from raw data, and build self-improving feedback loops with minimal human intervention.

Expanded Meaning of the Emerging AI-Spectroscopy Themes

Several emerging themes are developing based on the application of AI in spectroscopy (22-30).

• Self-driving spectroscopic laboratories, in theory, automate the full experimental cycle, from preparing samples and setting instrument conditions to acquiring, interpreting, and refining results. Using closed-loop reinforcement learning, the AI system experiments, evaluates its own output, and adjusts the next step automatically, similar to how autonomous vehicles learn through trial-and-error (3,22,23). This can accelerate reaction discovery, calibration development, and multi-day experiments that previously required continuous oversight.
• Multimodal fusion networks combine information from Raman, IR, mass spectrometry (MS), and nuclear magnetic resonance (NMR) to provide a more complete chemical fingerprint than any single analytical measurement technique (24,25). For general scientists, this means a single model can identify molecular structure, functional groups, isotopic patterns, and even trace impurities with greater confidence, much like merging multiple medical imaging scans to improve diagnostic accuracy.
• Physics-informed neural networks (PINNs) embed real chemical rules, such as vibrational symmetry, anharmonicity, and IR-Raman selection rules, directly into the training process (26,27). Instead of learning blindly from data, the model is physically grounded, reducing overfitting and improving interpretability. This bridges traditional theory with ML, giving results that are not only accurate but scientifically meaningful.
• Spectroscopy-generalist large language models (LLMs) extend beyond text; they can read spectra like a scientist reads a graph, label peaks, suggest band assignments, and even generate multivariate calibration strategies automatically (28,29). For researchers, this could be equivalent to having a virtual postdoc who can draft methods, troubleshoot models, or preprocess spectra with natural-language instructions.
• Generative diffusion models synthesize realistic spectra by learning the statistical structure of large spectral datasets (30). These synthetic spectra can be used to train models when experimental data are scarce, extend calibration ranges for rare analytes, and simulate instrument conditions otherwise difficult to access. This is particularly powerful for early-stage biomarker discovery, environmental monitoring, and low-concentration detection problems.

In essence, the new AI implementation describes the emergence of foundation-scale spectroscopy AI–systems that combine physics, ML, and generative modeling to operate flexibly across different instrumental techniques, a variety of sample types, and diverse scientific domains, reducing bottlenecks and democratizing high-level analytical insight.

Summary

AI in vibrational spectroscopy advanced dramatically in 2025. The field showed maturation from model-driven analysis into predictive, uncertainty-aware autonomous intelligence. Key achievements include deep-learning Raman microplastic detection, soil-carbon quantification using MIR CNNs, SERS-powered biomedical diagnostics, NIR explosive detection, and foundation-scale multi-sensor fusion frameworks (1–21). Future development will prioritize data and model explainability, hybrid physics combined with data modeling, and generalist spectral LLMs capable of zero-shot interpretation across modalities (1,5,20-30).

A generalist spectral LLM is a large language model trained not just on text, but also on many kinds of spectral data (Raman, IR, NIR, MS, NMR, UV–Vis, hyperspectral imaging, and so forth). Instead of being built for one task or one instrument type, it is designed to learn broad chemical interrelationships in a similar way that a human expert develops domain knowledge through years of experience.

The phrase zero-shot interpretation across modalities means the model can correctly interpret a spectrum it has never seen before without needing additional training examples. It can look at a new measurement, recognize patterns, assign peaks, and suggest the presence of likely compounds or functional groups on the first modeling attempt. It does this across multiple spectroscopy techniques, switching flexibly the way a scientist might move from Raman to NMR when solving a structural problem.

Bridging Bioscience, Environment, and Industry: Unified Trends

Examining the 2025 literature as a whole, a few convergent themes emerge.

• Multimodal data fusion: Many new applications combine different spectroscopic modalities (for example, NIR + Raman), or integrate spectral data with imaging, remote sensing, or temporal process data (for example, drone-based hyperspectral soil sensing; dual-sensor NIR-Raman spectroscopy in fermentation; SERS + imaging + ML for environmental toxins). This trend reflects aspirations to build holistic, context-aware spectral systems (1,6,12,19,20).
• Generative and foundation models: Deep-learning software platforms like Vib2Mol point toward a future in which spectral data can be used not only for classification or calibration, but also for generation (structure prediction), enabling autonomous discovery, materials design, and reaction monitoring (6).
• Explainability and trustworthiness: Software tools like SpecReX and QRF highlight a critical shift toward interpretable AI, uncertainty quantification, and regulatory-ready spectral models (4,5).
• Real-time, in situ, and non-destructive workflows: Applications in wearable sensors, industrial bioprocess monitoring, remote environmental monitoring, and non-invasive diagnostics suggest that AI-powered spectroscopy is rapidly moving from the lab to the field, clinic, and factory floor.
• Scalability and accessibility: By automating feature extraction and leveraging small-sample learning (for example, CNN-based self-supervised learning in NIR (29), AI is lowering the barrier for deploying spectroscopic analysis in resource-limited environments and among non-expert users.

Table I: Summary of reported 2025 AI-spectroscopy advances

Future Outlook: Toward Autonomous, Interpretable, Foundation-Model Spectroscopy

The momentum built in 2025 suggests that AI-driven vibrational spectroscopy is on the cusp of entering a new phase of maturity. Several converging trends point toward a future in which spectral analysis becomes autonomous, interpretable, and deeply integrated into scientific, industrial, and clinical workflows.

• Foundational Models & Synthetic Data: As argued by Westermayr and Marquetand (6), and echoed in the survey on SpectraML (20), large-scale pretraining, synthetic-spectrum generation, few- and zero-shot learning, and foundation-model architectures (graph nets, transformers) could enable models that generalize across molecular classes, spectroscopic modalities, and even measurement conditions.
• Physics-Informed Neural Networks: By embedding chemical and physical constraints (conservation laws, vibrational mode coupling) into neural network architectures, future models may overcome limitations of purely data-driven methods , improving generalizability, interpretability, and robustness in complex environments such as biological matrices or industrial reactors (1,6).
• Multimodal & Multiscale Integration: Combining vibrational spectroscopy with spectrometric methods (for example, MS), imaging, chromatography, remote sensing, and even time-series process data could enable holistic spectral systems—unifying structural, compositional, spatial, and temporal information in one AI-driven framework (1,20).
• Explainable, Regulatory-Ready Spectroscopy: As explainable AI tools (SpecReX) mature, and as uncertainty-aware models (QRF) become more common, AI-spectroscopy can gain broader acceptance in regulated domains such as medicine, environmental monitoring, and pharmaceuticals.
• Real-Time, Embedded, and Wearable Systems: With advances in wearable SERS sensors (17), drone- and imaging-based environmental spectroscopy (10,14), and in-line bioprocess monitoring (19), future workflows may rely on AI-driven spectral analytics embedded in IoT or edge devices enabling continuous, real-time, non-destructive sensing at scale.

In short: 2025 may well be remembered as the year when vibrational spectroscopy matured from an expert-driven, lab-bound art into an intelligent, autonomous, and scalable analytical data science.

Conclusion

The advancements of 2025 demonstrate that AI has begun to realize its promise in reshaping vibrational spectroscopy. From humble beginnings in classical chemometrics, AI now supports deep learning, generative modeling, uncertainty quantification, and spectrum-to-structure inference. Real-world applications in agriculture, environmental sensing, industrial biotechnology, and medicine illustrate the breadth and depth of this transformation.

Still, challenges remain: data standardization, domain-specific interpretability, regulatory acceptance, and the need for robust, physics-aware models. But with the emergence of explainable AI tools, foundation-model frameworks, and open-source platforms like SpectraML and related toolkits, the path forward is becoming clearer.

If 2025 marks the beginning of “intelligent spectroscopy,” the coming years are likely to see its full maturation as a future in which spectral data flows through autonomous pipelines, delivering actionable chemical, biological, and materials insights in real time, with transparency, reproducibility, and at scale.

The big question will be, “Does it all work as expected?”

References

(1) Workman, J., Jr. Recent Research in Chemometrics and AI for Spectroscopy, Part II: Emerging Applications, Explainable AI, and Future Trends. Spectroscopy 2025, Online, November 4. Available at: https://www.spectroscopyonline.com/view/recent-research-in-chemometrics-and-ai-for-spectroscopy-part-ii-emerging-applications-explainable-ai-and-future-trends (accessed 2025-12-2).

(2) Workman, J., Jr. Recent Research in Chemometrics and AI for Spectroscopy, Part I: Foundations, Definitions, and the Integration of Artificial Intelligence in Chemometric Analysis. Spectroscopy 2025, Online, November 3. Available at: https://www.spectroscopyonline.com/view/recent-research-in-chemometrics-and-ai-for-spectroscopy-part-i-foundations-definitions-and-the-integration-of-artificial-intelligence-in-chemometric-analysis (accessed 2025-12-2).

(3) Workman, J., Jr. Humans and Machines Unite Using Symbiotic AI to Transform Polymer Chemistry and Spectroscopy. Spectroscopy 2025, Online, October 15, 2025. Available at: https://www.spectroscopyonline.com/view/humans-and-machines-unite-using-symbiotic-ai-to-transform-polymer-chemistry-and-spectroscopy (accessed 2025-12-03).

(4) Workman, J., Jr. Smarter Spectroscopy With a New Machine Learning Approach to Estimate Prediction Uncertainty. Spectroscopy 2025, Online, August 27, 2025. Available at: https://www.spectroscopyonline.com/view/smarter-spectroscopy-with-a-new-machine-learning-approach-to-estimate-prediction-uncertainty (accessed 2025-12-03).

(5) Workman, J., Jr. How Is Explainable AI Transforming Spectroscopy? Spectroscopy 2025, Online, October 8. Available at: https://www.spectroscopyonline.com/view/how-is-explainable-ai-transforming-spectroscopy- (accessed 2025-10-08).

(6) Workman, J. Jr.; Mark, H. From Classical Regression to AI and Beyond: The Chronicles of Calibration in Spectroscopy: Part II. Spectroscopy 2025, 40 (7), 6–10. DOI: 10.56530/spectroscopy.fc1076p9

(7) Workman, J., Jr. Machine Learning Enhanced Raman Spectroscopy for Microplastics Detection in Environmental Samples: A Practical Tutorial. Spectroscopy 2025, Online, August 18, 2025. Available at: https://www.spectroscopyonline.com/view/machine-learning-enhanced-raman-spectroscopy-for-microplastics-detection-in-environmental-samples-a-practical-tutorial (accessed 2025-10-10).

(8) Workman, J., Jr. AI Boosts SERS for Next Generation Biomedical Breakthroughs. Spectroscopy 2025, Online, July 2, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-boosts-sers-for-next-generation-biomedical-breakthroughs (accessed 2025-12-03).

(9) Workman, J., Jr. Artificial Intelligence Accelerates Molecular Vibration Analysis, Study Finds. Spectroscopy 2025, Online, July 1, 2025.
Available at: https://www.spectroscopyonline.com/view/artificial-intelligence-accelerates-molecular-vibration-analysis-study-finds (accessed 2025-12-03).

(10) Workman, J., Jr. AI and Infrared Light Team Up to Advance Soil Carbon Monitoring. Spectroscopy 2025, Online, May 19, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-and-infrared-light-team-up-to-advance-soil-carbon-monitoring (accessed 2025-12-03).

(11) Workman, J., Jr. AI-Driven Raman Spectroscopy Paves the Way for Precision Cancer Immunotherapy. Spectroscopy 2025, Online, April 15, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-driven-raman-spectroscopy-paves-the-way-for-precision-cancer-immunotherapy (accessed 2025-12-03).

(12) Workman, J., Jr. Advancing Deep Soil Moisture Monitoring with AI-Powered Spectroscopy Drones. Spectroscopy 2025, Online, July 7, 2025.
Available at: https://www.spectroscopyonline.com/view/advancing-deep-soil-moisture-monitoring-with-ai-powered-spectroscopy-drones (accessed 2025-12-03).

(13) Workman, J., Jr. AI and Satellite Spectroscopy Team Up to Monitor Urban River Pollution in China. Spectroscopy 2025, Online, April 30, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-and-satellite-spectroscopy-team-up-to-monitor-urban-river-pollution-in-china (accessed 2025-12-03).

(14) Workman, J., Jr. Smart Farming Using AI, IoT, and Remote Sensing. Spectroscopy 2025, Online, March 4, 2025. Available at: https://www.spectroscopyonline.com/view/smart-farming-using-ai-iot-and-remote-sensing (accessed 2025-12-03).

(15) Workman, J., Jr. New Technique Combines Raman Spectroscopy and AI to Accurately Detect Microplastics in Water, Spectroscopy 2025, Online, August 19, 2025 Available at: https://www.spectroscopyonline.com/view/new-technique-combines-raman-spectroscopy-and-ai-to-accurately-detect-microplastics-in-water (accessed 2025-12-03).

(16) Workman, J., Jr. Mini-Tutorial: Raman Fingerprinting and Machine Learning Classification of Pesticides Using a Custom 785 nm Instrument. Spectroscopy 2025, Online, August 4, 2025. Available at: https://www.spectroscopyonline.com/view/mini-tutorial-raman-fingerprinting-and-machine-learning-classification-of-pesticides-using-a-custom-785-nm-instrument (accessed 2025-10-10).

(17) Workman, J., Jr. The Rise of Smart Skin Using AI-Powered SERS Wearable Sensors for Real-Time Health Monitoring. Spectroscopy 2025, Online, May 5, 2025. Available at: https://www.spectroscopyonline.com/view/the-rise-of-smart-skin-using-ai-powered-sers-wearable-sensors-for-real-time-health-monitoring (accessed 2025-12-03).

(18) Workman, J., Jr. AI-Powered SERS Spectroscopy Breakthrough Boosts Safety of Medicinal Food Products. Spectroscopy 2025, Online, April 16, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-powered-sers-spectroscopy-breakthrough-boosts-safety-of-medicinal-food-products (accessed 2025-12-03).

(19) Workman, J., Jr. AI and Dual-Sensor Spectroscopy Supercharge Antibiotic Fermentation. Spectroscopy 2025, Online, June 30, 2025.Available at: https://www.spectroscopyonline.com/view/ai-and-dual-sensor-spectroscopy-supercharge-antibiotic-fermentation (accessed 2025-12-03).

(20) Workman, J., Jr. AI-Powered Near-Infrared Imaging Remotely Identifies Explosives. Spectroscopy 2025, Online, June 11, 2025.Available at: https://www.spectroscopyonline.com/view/ai-powered-near-infrared-imaging-remotely-identifies-explosives (accessed 2025-12-03).

(21) Workman, J., Jr. AI Shakes Up Spectroscopy as New Tools Reveal the Secret Life of Molecules. Spectroscopy 2025, Online, April 14, 2025.
Available at: https://www.spectroscopyonline.com/view/ai-shakes-up-spectroscopy-as-new-tools-reveal-the-secret-life-of-molecules (accessed 2025-12-03).

(22) Workman, J., Jr.; Mark, H. From Classical Regression to AI and Beyond: The Chronicles of Calibration in Spectroscopy: Part I. Spectroscopy 2025, 40 (2), 13–18. DOI: 10.56530/spectroscopy.pu3090t7

(23) Workman, J., Jr. Advancing Near-Infrared Spectroscopy and Machine Learning for Personalized Medicine. Spectroscopy 2025, Online, February 12, 2025. Available at: https://www.spectroscopyonline.com/view/advancing-near-infrared-spectroscopy-and-machine-learning-for-personalized-medicine (accessed 2025-12-03).

(24) Workman, J., Jr. New Method for Detecting Fentanyl in Human Nails Using ATR FT-IR and Machine Learning. Spectroscopy 2025, Online, February 11, 2025. Available at: https://www.spectroscopyonline.com/view/new-method-for-detecting-fentanyl-in-human-nails-using-atr-ft-ir-and-machine-learning (accessed 2025-12-03).

(25) Workman, J., Jr. New AI-Powered Raman Spectroscopy Method Enables Rapid Drug Detection in Blood. Spectroscopy 2025, Online, February 10, 2025.
Available at: https://www.spectroscopyonline.com/view/new-ai-powered-raman-spectroscopy-method-enables-rapid-drug-detection-in-blood (accessed 2025-12-03).

(26) Workman, J., Jr. Blood-Glucose Testing: AI and FT-IR Claim Improved Accuracy to 98.8%. Spectroscopy 2025, Online, February 3, 2025.
Available at: https://www.spectroscopyonline.com/view/blood-glucose-testing-ai-and-ft-ir-claim-improved-accuracy-to-98-8- (accessed 2025-12-03).

(27) Workman, J., Jr. NIR Spectroscopy with AI Proves to be a Powerful Combination for Tea Classification. Spectroscopy 2025, Online, January 29, 2025. Available at: https://www.spectroscopyonline.com/view/nir-spectroscopy-with-ai-proves-to-be-a-powerful-combination-for-tea-classification (accessed 2025-12-03).

(28) Workman, J., Jr. Optical Insights into Fibrotic Livers: Applications of Near-Infrared Spectroscopy and Machine Learning. Spectroscopy 2025, Online, January 28, 2025. Available at: https://www.spectroscopyonline.com/view/optical-insights-into-fibrotic-livers-applications-of-near-infrared-spectroscopy-and-machine-learning (accessed 2025-12-03).

(29) Workman, J., Jr. The Marriage of Near-Infrared Spectroscopy with AI: The Small-Sample Breakthrough. Spectroscopy 2025, Online, January 27, 2025.
Available at: https://www.spectroscopyonline.com/view/the-marriage-of-near-infrared-spectroscopy-with-ai-the-small-sample-breakthrough (accessed 2025-12-03).

(30) Workman, J., Jr. Machine Learning-Enhanced SERS Technology Advances Cancer Detection. Spectroscopy 2025, Online, January 13, 2025.
Available at: https://www.spectroscopyonline.com/view/machine-learning-enhanced-sers-technology-advances-cancer-detection (accessed 2025-12-03).