Hi!👋, I'm Anirban
and I'm
PhD in Machine Learning and Cosmology at Institut d'Astrophysique de Paris, CNRS & Sorbonne Université
ML | Stats | CV | NLP | Gen-AI
ContactBiography
Introduction
Anirban Bairagi is a doctoral fellow at the Institut d’Astrophysique de Paris, funded by the Simons Foundation, working under the supervision of Prof. Benjamin Wandelt.
His research lies at the intersection of cosmology, astrophysics, statistics, and machine learning.
He is a member of the international Euclid Consortium and an active member of the Simons Collaboration on Learning the Universe.
Born in 1998 and raised in a small town near Calcutta, India, Anirban earned his Bachelor’s and Master’s degrees from the Indian Institute of Technology Kharagpur in 2022.
During his stay at IIT Kharagpur, he began his research journey under the mentorship of Prof. Sayan Kar, Prof. Somnath Bharadwaj, and Prof. Jyotirmoy Bhattacharya, in the field of general relativity and cosmology.
After graduation, he briefly worked as a consultant at TCG Digital before beginning his Ph.D. in Paris.
Anirban has been awarded several prestigious fellowships, including the Caltech LIGO SURF, MITACS Globalink Research Fellowship, and the Shastri Indo-Canadian Fellowship, recognizing his dedication to advancing astrophysical science.
His current research focuses on the theoretical and computational aspects of cosmology, especially the Large-Scale Structure of the Universe and its primordial origins.
His work integrates analytical theory, numerical simulations, and machine learning techniques to address fundamental questions related to cosmic structure formation and data-driven statistical inference from upcoming astronomical surveys.
Expertise
Technical skillsCoding Languages
Python
90%
Cython
70%
Mathematica
90%
C
75%
SQL
60%
Artificial Intelligence
Machine Learning
90%
Deep Learning
90%
Convolutional Neural Network
90%
Generative AI
70%
Natural Language Processing
70%
Frameworks
Pytorch
90%
Tensorflow
80%
Keras
85%
Cuda
70%
Web development
HTML
70%
CSS
60%
OS
Linux
70%
Unix
70%
Windows
80%
Tools
Weights & Biases
80%
Git
75%
Cmake
70%
Matlab
75%
Arduino
75%
Certifications
Coursera
Neural Networks and Deep Learning
Preview
Convolutional Neural Networks
Preview
Diffusion Models
Preview
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
Preview
Structuring Machine Learning Projects
Preview
AI for Medical Diagnosis
Preview
Anomaly Detection
PreviewPortfolio
Most recent work
PatchNet: GPU and Simulations are not the limitation anymore for Cosmological Field-level Inference
Maximizing Fisher information content by combining information from different scales hierarchically.
Draft in prep.
How many simulations do we need for simulation-based inference?
Neural scaling law that predicts number of training-simulations required to get optimally informative neural summary.
Read more
LtU: Cosmological contraints from Wavelet coefficients
Leading the wavelet analysis for Simons collaboration on Learning the Universe. Used an ensemble of normalizing flows to train our model on fastPM simulations and validating on N-body simulations to confirm robustness before we apply it on real surveys.
Ongoing
Point-set clustering correction using a displacement potential
A method to correct small scale structures analytically in cheap Particle-Mesh like simulations to make it look like Nbody.
Draft in prep.
M.Sc. Thesis: Gravitational Wave Memory effect and Geodesic Congruences
Memory observables, Jacobi propagators, B-tensor formalism, focusing time calculation for triangular pulse in exact radiative spacetime.
Read more
Caltech SURF: LIGO Laser Beam Tracking
Precise measurement of laser beam position in Gravitational Wave detector due to movement of the mirrors using Convolutional Neural Networks with sub-pixel accuracy (<40 micron).
Read more
Gravitational Waves Detection and Glitch Classification using Convolutional Neural Network
Real-time detection of GW signals to be prepared for the detection of electro-magnetic counterparts, and glitch identification to mitigate the noise from future observations.
Read more