I am a biophysicist and AI researcher specialising in quantitative fluorescence microscopy,
deep learning, and scientific image analysis. During my PhD at Aix–Marseille University,
I developed U-Net–based density map prediction frameworks to count bacteriophages under TIRF
microscopy and estimate antigen–antibody binding kinetics. My work combines interpretable AI
with physics-driven validation using Poisson statistics, Langmuir models, and controlled
synthetic data.
Beyond microscopy, I work across Raman and FTIR spectroscopy, digital pathology, and multimodal
biological data analysis. I enjoy designing reproducible AI pipelines, building clear scientific
visualisations, and crafting computational tools that reveal the hidden patterns linking physics,
biology, and machine intelligence.
My long-term goal is to develop reliable, interpretable AI systems for biological imaging and
real-world scientific applications, bridging fundamental science with practical impact.
“The universe is full of hidden structure. What we call randomness is often a pattern awaiting discovery.
When we shift our perspective and pursue truth with clarity, the world reveals itself in unexpected ways.”
Research Interests
Deep learning for quantitative fluorescence microscopy and TIRF imaging
Density map prediction, particle counting, and weakly supervised learning
Physics-informed modelling: Poisson statistics, Langmuir kinetics, and mechanistic validation
Domain adaptation, synthetic data generation, and generalization across imaging conditions
AI for digital pathology, microscopy, and spatial biology
Multimodal learning: integrating microscopy, spectroscopy, and biophysical signals
Raman & FTIR spectroscopy for food quality, authentication, and chemical analysis
Interpretable and reliable AI systems for scientific and medical imaging
Reproducible scientific pipelines, HPC computation, and scientific MLOps
Selected Publications
Detection of hotspots in fluorescence imaging of yeast cells for neurodegeneration studies S. N. Chandrasekar, A. Rao, S. Venketesh, R. R. Sarma IEEE ICAECT, 2021. DOI
Deep learning-based detection of urethral stricture: segmentation & classification G. Nikhil, K. Udaya, S. N. Chandrasekar, et al. medRxiv preprint, 2024. DOI
Raman/FTIR-PCA frameworks for oil quality monitoring in a food matrix J. Gupta, A. Shaw, S. V. Muthukumar, S. N. Chandrasekar, et al. Food Chemistry, under review, 2025. Preprint
AI-based classification of edible oils in food matrix using Raman spectral signatures J. Gupta, A. Shaw, S. V. Muthukumar, S. N. Chandrasekar, et al. Food Control, under review, 2025.
Comparative analysis of CNN, VGG16, and ResNet-50 on spatial, FFT, and wavelet representations of MNIST and Fashion MNIST S. N. Chandrasekar, D. L. N. Kallepalli IJCVSP, under review, 2025.
Projects
My work combines deep learning, scientific image analysis, and physics-informed modeling to solve real
problems in microscopy, spectroscopy, and biomedical imaging. Below is a curated selection of projects that
reflect my interdisciplinary expertise, reproducible research practices, and interest in building reliable,
interpretable AI systems for science, medicine, and real-world applications.
PhD Project : Deep Learning Based Counting and Kinetic Analysis of Bacteriophages (Aix–Marseille University)
Developed U-Net based variants for precise density map prediction to identify and count single bacteriophages imaged under TIRF fluorescence microscopy.
Designed physics-informed workflows integrating Poisson statistics and Langmuir binding models to validate model prediction along with estimating antigen–antibody affinity
directly from microscopy data. Built fully reproducible AI pipelines on HPC clusters, including automated preprocessing,
augmentation, training, model evaluation, and statistical validation. This work bridges biophysics, computer vision,
and Interpretable week supervision Deep learning model by linking pixel-level predictions to real molecular kinetics.
Skills: Python, TensorFlow, PyTorch, Model architectures, Density map estimation, Statistical modelling, HPC/SLURM, Domain Adoptation, Multi-task optimization, TIRF microscopy.
AI for Raman & FTIR Spectroscopy : Food Quality, Oil Classification, and Chemical Fingerprinting (Dept. Food & Nutritional Sciences, SSSIHL)
Developement of machine learning pipelines for Raman and FTIR spectroscopy to Identify and authenticate edible oils and analyze processed food matrices.
Implemented PCA, feature engineering, ratiometric analysis, and classical ML (LR, KNN, SVM, Random Forest, XGBoost, LightGBM).
Explored multi-domain representations (raw spectra, preprocessed spectra, Band bining). Decomposition of multi-ingredient spectral interact (NNLS) and contributed to manuscripts and
experimental data analysis. This work demonstrates how AI can extract chemical insights from complex spectral signatures and improves food quality monitoring using low-cost spectroscopy.
Skills: Spectral analysis, Data Exploration, PCA, sklearn, XGBoost/LightGBM, Feature extraction, Data visualization, Signal processing.
Urethral Stricture Detection, Segmentation & Classification using Retrograde Urethrogram (SSSIHL & SSSIHMS Collaboration)
Served as an independent collaborator providing methodological guidance, mentoring, and image-analysis support for a clinical
research project between the Department of Urology (SSSIHMS) and the Department of Physics (SSSIHL). The study focused on analyzing
Retrograde Urethrogram (RUG) X-ray fluoroscopic images to identify and segment urethral strictures, quantify stricture geometry, and
classify clinical severity. I contributed to the design of preprocessing workflows (contrast enhancement, noise reduction,
anatomical masking), classical CV feature extraction, and early deep-learning prototyping. This collaboration supported a medRxiv
preprint (doi:10.1101/2024.10.16.24315644) and represents a step toward integrating AI-assisted interpretation into RUG-based
urological diagnostics.
Skills: RUG fluoroscopic image analysis, Segmentation & classification workflows, Classical CV, Python/OpenCV, Research collaboration, Mentoring, Medical imaging.
Medical Device Interface Engineering — UI/UX & Embedded Integration (Phoenix Medical Systems)
Designed and developed Qt/QML-based user interfaces and control panels for neonatal medical devices including infant incubators and
ventilators. Built responsive UI components, alarm visualization modules, and interaction workflows aligned with clinical usability and
safety requirements. Developed real-time communication interfaces between the embedded hardware and the front-end, ensuring reliable
data exchange, fault handling, and device-state visualization. Collaborated with biomedical engineers and clinicians to refine
interaction logic, improve ergonomics, and support regulatory-grade software development practices within a medical-device pipeline.
Skills: Qt/QML, Embedded UI development, Hardware–software integration, Serial communication interfaces, Medical device workflows,
Human-centered design, UI/UX for clinical systems.
H&E Slide Analysis for Lymphoma Detection (Grey Scientific Labs)
Developed OpenCV-based image analysis pipelines for identifying lymphoid follicles in H&E-stained slides as part of an
early-stage lymphoma detection workflow. Worked with whole-slide image (WSI) data structures, designed modular preprocessing
components, and performed morphological & texture-based feature extraction.
Learned and applied production-grade practices including Git versioning, reporting standards, structured experiment
tracking, and reproducible algorithm development—building foundations essential for deployment-ready digital pathology tools.
Skills: OpenCV, Morphological operations, WSI processing, Color normalization, Git & version control, Reproducible pipelines, Biomedical image analysis.
Master’s Thesis : Computer-Aided Diagnosis of Renal Biopsy
Designed and implemented a Python-based computer-aided diagnosis system for quantitative analysis of renal biopsy microscopy images.
Developed a complete processing pipeline combining deep learning and classical computer-vision methods to segment glomeruli, extract structural
and textural biomarkers, and classify tissue health indicators relevant to renal pathology. The workflow incorporated morphological analysis,
color deconvolution, feature engineering, and patch-based CNNs to support clinical interpretation. Worked closely with subject-matter
experts to validate visual outputs and iteratively refine the diagnostic criteria.
Skills: Renal biopsy microscopy, Glomeruli segmentation, OpenCV, PyTorch, TensorFlow, Feature extraction, Classical ML, Biomedical image interpretation, Morphological analysis.
Fluorescence Imaging of Yeast Cells for Neurodegenerative Disease Research (Dept. Biosciences, SSSIHL Collaboration)
Developed an image-analysis pipeline to quantify molecular hotspots in fluorescence microscopy images of yeast cells exposed to
neurodegeneration-related chemical stressors (associated with Parkinson’s and Huntington’s disease models). Implemented preprocessing,
segmentation, and hotspot-detection algorithms using OpenCV to measure spatial intensity patterns and cellular responses.
Generated statistically interpretable visualizations that supported biological insight into stress-induced protein aggregation.
This work led to a peer-reviewed IEEE conference publication (ICAECT 2021).
Skills: Fluorescence microscopy, Hotspot detection, Python/OpenCV, Image segmentation, Scientific visualization, Data-driven biological analysis, Research writing.
31-Band Audio Equalizer — Digital Signal Processing
Developed a Python-based 31-band graphic equalizer using discrete Fourier transform (FFT) techniques to analyse and modify audio signals
in the frequency domain. Implemented spectral decomposition, band-pass isolation, amplitude scaling, and inverse FFT reconstruction to
enable fine-resolution control across standard equalizer frequency bands. Designed a complete offline processing pipeline that supports
experimentation with spectral shaping, filter design, and acoustic enhancement. This project explores practical applications of digital
signal processing, from frequency-domain manipulation to perceptual audio modification.
Skills: FFT & spectral analysis, Digital signal processing, Python (NumPy/SciPy), Frequency-band filtering, Audio manipulation, Algorithm development.
Scientific Content Automation & Infrastructure Development for Virtual Labs (Department of Physics, SSSIHL)
Contributed to the development of the Virtual Labs platform by building automation pipelines and backend infrastructure for
undergraduate physics experiments. Designed Python- and SQL-based tools to convert scientific experiment documentation into standardized,
web-ready formats, enabling efficient and reproducible deployment of multiple physics laboratory modules. Implemented workflows to manage
HTML/JavaScript content, generated structured documentation templates, and integrated experiments with IIS- and Kiwix-based servers for
both online and offline access. Assisted in front-end refinement, asset optimization, and end-to-end content delivery processes that
strengthened Technology Enabled Learning (TEL) within the Department of Physics.
Skills: Python automation, SQL, HTML/CSS/JavaScript, C#, Backend–frontend integration, IIS server deployment, Kiwix/ZIM packaging,
Educational technology (EdTech), Workflow and documentation pipeline design.
Teaching, Mentoring & Talks
I contribute to teaching and mentoring in AI, quantitative microscopy, and scientific computing.
Over the years, I have guided Master's students, interns, and junior researchers on
deep learning model design, image analysis workflows, reproducible research practices,
and scientific communication.
Mentoring & Supervision
(Named students can be added as you mentor more people.)
Mr. Nikhil Gurung —
Lead collaborator on deep learning–based urethral stricture detection.
Dr. Ashwin Shekar P —
Clinical collaborator for urethral stricture dataset & image analysis.
Dr. L Udaya Kumar —
Clinical collaborator for urethral stricture dataset & image analysis.
Talks & Presentations
(You can expand this list as your speaking engagements grow.)
16 September 2025 — Deep Learning-Based Counting and Kinetic Analysis of Bacteriophages from TIRF Microscopy,
PhD Defense Seminar, Aix–Marseille University (In-person).
13 February 2024 — Internal Research Presentation,
Laboratoire Adhésion Inflammation (LAI), Luminy, Marseille.
4 June 2024 — Internal Research Presentation,
Laboratoire Adhésion Inflammation (LAI), Luminy, Marseille.
28–31 October 2024 — Poster Presentation,
JMC – French Condensed Matter Days 2024, France.
12 November 2024 — Poster Presentation,
“AI and Physical Sciences @ AMU”, Aix–Marseille University.
2021 — Detection of hotspots in fluorescence imaging of yeast cells for neurodegeneration studies,
IEEE ICAECT Conference (Online).
202X — Placeholder Talk Title,
Placeholder Conference or Seminar, City / Online.
News
16 September 2025 · PhD Defense
Defended my PhD thesis in Biophysics at Aix–Marseille University
I successfully defended my PhD thesis on deep learning-based counting and binding kinetics of
bacteriophages imaged by TIRF microscopy. The work combines U-Net based density map prediction,
physics-informed validation with Poisson statistics, and Langmuir kinetic modelling of antigen–antibody
interactions.
It was a special day shared with my supervisors, colleagues at LAI and LIS, and my family following the
journey from Chennai to Marseille. The defense highlighted both the technical contributions and the
interdisciplinary bridge between physics, AI, and immunology.
After the defense at Aix–Marseille University with the jury and supervisors.
Our collaborative work on Raman and FTIR spectroscopy for edible oil quality monitoring in chips and pure oils
has led to two manuscripts, currently under review. We developed PCA- and ML-based frameworks to track changes
in oil quality under heating and to classify oils from complex food matrices.
My contribution focused on building robust Python pipelines, PCA/ML analysis, and clear visualization of
ratiometric spectral features, with an emphasis on interpretability for food scientists and engineers.
Raman/FTIR spectral analysis workflow for edible oil quality monitoring.
June 2025 · Methods Paper
Wavelet and FFT domain analysis of CNN, VGG16 & ResNet-50
I submitted a methodological paper that compares CNN, VGG16, and ResNet-50 across spatial, FFT, and wavelet
representations of MNIST and Fashion MNIST. The study investigates how different frequency-domain
representations affect robustness, generalization, and behaviour under partial test set removal.
This work connects my interests in classical signal processing with modern deep learning and is intended to
be a resource for researchers exploring frequency-aware architectures and data representations.
I am happy to connect about collaborations, research discussions, or
opportunities in AI, biophysics, digital pathology, spectroscopy, and
scientific imaging.
