About Me
Hi, my name’s Ehsan and I’m a Machine Learning Researcher. I have applied Deep Learning in areas including but not limited to High Energy Physics, Molecule Generation, Aerial Imaging and Medical Imaging.
I am most skilled in: Deep Learning and Computer Vision
Projects
Using Deep Generative Modelling to detect Anomalies in data coming from Large Hadron Collider.
This project was created using Pytorch. It implements Masked AutoRegressive Flows model for modelling the data distribution of events from LHC. The project allows for sampling new data points from the distribution as well as inferring whether a data point belongs to the background or signal (anomaly) distribution.
Detecting Vehicles from satellite images.
This project explores the performance of different object detection models (YOLO-based, RCNN-based and RetinaNet (FPN-Based)) on aerial vehicle detection. “SIMD” dataset was used for this purpose. It was built using Tensorflow.
Nuclear Segmentation in digital microscopic tissues via Deep Learning based Computer Vision techniques.
Using SOTA models in semantic segmentation (DeepLab, SegNet and UNet) nuclear segmentation was carried out. MonuSeg dataset was used. Built using Tensorflow.
Experience
Applying Machine Learning in High Energy Physics.
Machine Learning has showed potential in various field, thus it only makes sense to apply it for High Energy Physics. During my work here, I applied Machine Learning for various tasks on Large Hadron Collider Dataset.
Freelance Project
Machine Learning Researcher
2020
Applying Machine Learning for molecule generation.
Generating Molecules using Deep Learning requires knowledge of Generative Modelling and Natural Language Processing. Here, I applied Variational AutoEncoders for generating sequences of atoms that composed a molecule. This also allowed for latent walk, where new molecules could be generated.
Education
National University of Sciences and Technology
BE Software Engineering
2015 - 2019
Ranked # 1 in Pakistan. Worldwide it is ranked at 143 for Computer Science while overall its position is 358.
Software Engineering degree is an intensive program where the student is exposed to various real-world problems. Collaboration with various industries is also carried out.
National University of Sciences and Technology
MS Computer Science
2019 - 2021
Ranked # 1 in Pakistan. Worldwide it is ranked at 143 for Computer Science while overall its position is 358.
Most of my degree is related to Deep Learning and Computer Vision. My thesis was in Vision based TV Analytics where I used principles from Software Engineering to apply Metric Learning to Live Video Streams from YouTube.