Hi, I am a physicist by education and Computational Biologist/data scientist by profession, having 6+ years of experience at R&D as a Sr. Algorithm Scientist/Optics Physicist at Thermo Fisher Scientific, a leading Electron Microscpy technology company.
In short 'About me'
I love solving data driven problems using modern Machine learning algorithms, with a strong inclination towards problems in life sciences, especially, in medical and healthcare domain.
To pursue my interest, recently I joined as a Postdoc in a computational structural biology lab with in-house Cryo-EM imaging facility. I have been working on different problems, applying my skills in machine learning and statistical techniques. Along this path I have developed strong expertise in various machine learning methods as well as handling variety of data types, such as 1D signals, 2D images, 3D volume datasets, text-data, and other multi-dimensional data sets (e.g. xyz, time, energy...). Here, I am posting some of the interesting problems that I have worked on as hobby. My recent published (and unpublished, hobby) projects can be found on my GUthub page, and on Linkedin page, linked below.
In my spare time I love running and cycling in the woods along the country side roads.
My professional background
Center for Innovation Competence (ZIK), Martin-Luther-University, Halle(Saale) Germany | 2021
Advanced Technology group, R&D, Thermo Fisher Scientific, Eindhoven, Netherlands | 2014-2020.
Max Planck Institute for Plasma Physics, Munich, Germany, and
Culham Science Centre, Oxford, UK | 2013
Technical University of Munich, Munich, Germany,
International Max Planck Reseach School, Munich, Germany | 2009 - 2013.
Inter University Center for Astronomy and Astrophysics (IUCAA), Pune and
National Center for Radio Astrophysics, Pune, India | 2008.
Something that is never over...
2020
University of Edinburgh, UC San Diego
Coursera.org/EdX.orgMedical image analysis, Natural language processing (LSTM), processing DNA sequence, finding motif.
2019-2020
DeepLearning.ai
Coursera.orgMachine learning methods, Neural Networks, TensorFlow, Convolutional Neural Networks, Transfer learning and building customized models for real life problems.
2017
Stanford University
Coursera.orgMathematical theory and formulation of dimensionality reduction (PCA/ICA), Regression, clustering and Gradient descent.
2009 -2013
Technical Univeristy of Munich, Munich
GermanyPlasma Physics, Python, spectral analysis, experimental physics, X-ray detector, radiation-matter interaction physics.
2006 - 2008
University of Pune,
IndiaCore subjects include physics, mathematical physics, and astrophysics.
Selected problems I enjoyed working on...
A simple computer-vision based algorithm to detect cell (spots-like) signals from microscopy images in Life Sciences. Contains simulated imaging-data with varying number of cells and intensities, with corresponding statistical analysis.
View here » Github »It is a Python Jupyter Notebook based simple GUI; a statistical tool for investigating a protein-protein interaction interface from multiple protein structures. Full article:
View here » Github »A dashboard tool to perform full Bayesian Parameter Estimation, a robust approach in quantifying a statistically significant difference or similarity between any two sets of measurements.
View here » Github »A Web-app based on Plotly Dashboard, to automatically pull recent workouts from cloud, recorded by Polar sport watch, and provide detailed insights over the progress (or not!).
View here » Github »Generated >1000 views on kaggle.com, making it one of the most popular notebooks, on the CNN based model to classify images as cats or dogs, WITHOUT transfer learning techniques.
View here » Github »Inside CNN: how the filters look like? what does the output at different layers look like? Visualize the feature maps and filters.
View here » Github »A CNN based model is developed to identify if a patient is suffering from Pneumonia or not. Model developed from scratch i.e. without transfer learning, show 91% test accuracy.
View here » Github »Recommend books, using Co-similarity as well as user-book matrix factorization approach. Plus a quick search engine to find a book given some keywords.
View here » Github »By using Natural Language Processing (NLP) techniques and matrix factorization (PCA/LDA), thousands of text articles are processed to identify different clusters.
View here » Github »