Hi,
I am Soumee Mukherjee
Software Developer
and Researcher

Heritage Institute of Technology

Bachelor's thesis

Project: Studying Developer Interaction in Large-scale Software Development Ecosystems

• Working on two manuscripts and built simulations on Netlogo regarding the formation of developer networks in large scale software development eco-systems and compared it to existing software ecosystem like Eclipse, Android and Openstack.

• Worked on Java backend and decreased code runtime by employing SQL tuning and optimisations.

• Created an UI on Streamlit to parse complicated csv data. Github Link

• Skills: Netlogo, Java, Streamlit, SQL optimisation.

Indian Statistical Institute

Machine Learning Course

Winter Workshop on Advanced Machine Learning

Link

University of Calgary

Deep Learning Research Intern, MITACS fellow

Project: Computerised Tomography Perfusion of Stroke using Machine Learning

• Prepared the pipeline to organise raw patient data having CT Perfusion Data

• Denoised CT images using Conditional GANs and Fourier Bsed denoising techniques

• Estimated Cerebral Blood Flow paramteres, which shows the stroked region

• Skills: Shell scripts, Linux, Python, Time-series Data Analysis, DICOM and NIFTI files, Data Cleaning

Jadavpur University

Application Developer

Project: Denoising Images Using Binarization techniques and Deep Learning Models

• Published a paper on Binarization of Medical Images, using a Meta-Consensus Model that takes the consensus of local and global thresholding methods and Deep Learning Methods(like Unets) to determine whether a pixel is black or white. Model accuracy: 92%

• Using traditional graph-based algorithms for calculation of spine density, dendritic loops in medical images to be integrated on the software JYNIA

• Estimated Cerebral Blood Flow paramteres, which shows the stroked region

• Link: https://jynia.org/

Heritage Institute of Technology

Undergraduate Researcher

Project: Detection of Diabetes in Retinal Images using Supervised Algorithms

• Separated the co-ordinates of pixels which include retinal vessels(white) and background(dark), used sampling techniques to balance the data.

• Applied machine-learning models on the data for segmentation of blood vessels from Medical images using traditional supervised Machine Learning Algorithms, like K-Means Clustering, Support Vector Machines and Neural Network to determine accuracy of image segmentation. Model accuracy: 78%

• Skills: Sampling Techniques(SMOTE, ADASYN), Image Segmentation, Python, Supervised Unsupervised algorithms

Calcutta Electric Supply Corporation

Internship Trainee

• Working at the Security Operations Centre for the one of the largest grids to distribute power in the metropolitan city of Kolkata.

• Skills: Network Security