Soumee Mukherjee

Research

Learning from complex, real-world systems

Work across agent-based modelling, biomedical image analysis, and the reliability of information systems.

Fellowship

MITACS Globalink Research Internship

University of Calgary · Calgary, Canada · Jul 2022 - Oct 2022

Selected for a funded research internship from a global pool of 20,000 applicants.

Conducted deep-learning research on CT perfusion stroke analysis in an international research environment.

Research experience

Feb 2023 - Aug 2023

Kolkata, India

Heritage Institute of Technology

Emergent Structure in Large-Scale Software Ecosystems via Agent-Based Modelling

Bachelor's Thesis Researcher

Studied how local developer interactions produce global collaboration structures in large software ecosystems.

  • Built agent-based simulations of evolving developer collaboration networks.
  • Validated synthetic network behaviour against Eclipse, Android, and OpenStack ecosystems.
  • Produced two manuscripts and supporting analysis tools for large CSV corpora.
Technologies:
  • NetLogo
  • Java
  • Streamlit
  • SQL
  • Complex Networks
GitHubPreprint

Jul 2022 - Oct 2022

Calgary, Canada

University of Calgary

CT Perfusion Stroke Analysis using Machine Learning

Deep Learning Research Intern, MITACS Fellow

Investigated machine-learning workflows for preparing and analysing CT perfusion stroke data.

  • Prepared data pipelines for DICOM and NIfTI patient imaging data.
  • Applied conditional GAN and Fourier-based denoising techniques.
  • Estimated cerebral blood-flow parameters used to delineate stroke regions.
Technologies:
  • Python
  • Shell
  • Linux
  • DICOM
  • NIfTI
  • Conditional GANs

Jul 2021 - Jan 2023

Kolkata, India

Jadavpur University

Denoising Medical Images via Binarization and Deep Learning

Application Developer

Developed image-segmentation and analysis methods for dendritic spine microscopy.

  • Combined local and global thresholding with U-Net segmentation in a meta-consensus model.
  • Applied graph algorithms to spine-density and dendritic-loop measurements.
  • Contributed methods for integration into the JYNIA neuroscience software.
Technologies:
  • C++
  • Python
  • U-Net
  • Transfer Learning
  • Linux
JYNIASpringer chapter

Apr 2021 - Jul 2021

Kolkata, India

Heritage Institute of Technology

Diabetic Retinopathy Detection via Supervised Learning

Undergraduate Researcher

Compared classical and neural approaches to retinal blood-vessel segmentation.

  • Prepared pixel-level vessel and background data from retinal fundus images.
  • Compared K-Means, support-vector machines, and neural-network models.
  • Applied SMOTE and ADASYN to address class imbalance.
Technologies:
  • Python
  • Image Segmentation
  • K-Means
  • SVM
  • SMOTE
  • ADASYN

Publications

  1. 2025

    On Separation and Connection: A Multi-Systems Study of Developer Interaction using Agent-Based Models

    Soumee Mukherjee, Subhajit Datta, Subhashis Majumder

    Information and Software Technology · accepted

    DOI: 10.2139/ssrn.5734055

  2. 2023

    A Meta-consensus Strategy for Binarization of Dendritic Spines Images

    Shauvik Paul, Nirmal Das, Subhrabesh Dutta, Dipannita Banerjee, Soumee Mukherjee, Subhadip Basu

    Proceedings of International Conference on Data, Electronics and Computing, Algorithms for Intelligent Systems, Springer, pp. 269-278 · published

    DOI: 10.1007/978-981-99-1509-5_25