Aman Agrawal
Quant Researcher - HFT | CQF (in process) | Ex-Goldman Sachs, Natwest Markets | MS - Computer Science
Quant Researcher - HFT | CQF (in process) | Ex-Goldman Sachs, Natwest Markets | MS - Computer Science
Python, Java, Javascript, R, Bash, SQL, HTML/CSS
Tensorflow, Keras, Pytorch, SpaCY, OpenCV, R, IBM Watson, Scikit-learn, NLTK, Gensim, Numpy, Scipy, Pandas, OpenAI-Gym, Snorkel
Django, Flask, REST, GCP, AWS, Git, UI-Path, Tableau, MongoDB, RESTful services
Machine Learning, Natural Language Processing, Algorithms Design and Analysis, Data Structures, Computer Vision, Deep Learning, Data Analysis, Statistical Modelling, Image Processing, Object Oriented Programming (OOP), Reinforcement Learning, Agile Methodologies
Worked with the Metaphor Identification project team led by Dr. Shlomo Argamon, Chair of Computer Science Department
AskArchie, AI-Centre of Excellence, Digital Engineering Services
Capital Calculator Engine, Credit Risk Reporting, Enterprise Solutions
As an intern at Smart Joules, worked on two projects that were successfully deployed as daily utility tools at the startup
Extracts a network of a subset of Twitter users (100-500), and reports the analysis on the extracted data
Developed classification of Brexit-related tweets using weak social supervision with only 10% of the labeled data. Used Snorkel for weak supervision and added unlabeled tweets for classification in addition to labeled tweets. Improved classifier accuracy by nearly 15%.
Perform Fitting using Active Contours to detect edges of an object in an image.
Calibrate 3D World points with the 2D image points.
Create a unique artistic image by composing two images where the new image will have content of one image and style of the other image.
Implemented this research paper on Deep Reinforcement Learning. Implemented a convolution neural network for Deep Q-Network (DQN) and linearly annealed epsilon-greedy policy to train the model to play Breakout and Space Invaders on OpenAI-Gym environment.
Developed a music generator trained on Nottingham dataset using Recurrent Neural Networks (RNNs). In order to account for time-delays between each notes played, each RNN unit is coupled with a Restricted Boltzmann Machine (RBM). The implementation used Keras with Theano backend to train the model on the GPU. By the end of the trianing, the model could produce good sounding music.
Support Vector Machine Classifier primarily use two hyper-parameters which determine its performance. These are C and gamma. This project studied the use of Particle Swarm Optimization (PSO) to find the optimum values of these hyper-parameters in the search space. PSO is a kind of evolutionary algorithm inspired from observing how in nature, the birds flock to find their food.
AI-bot to play chess using basic Minimax algorithm and Alpha-beta Pruning. In order to deal with the combinatorial explosion, only a few levels deep tree is studied. Deployed this model to an Android Application for easier evaluation. The bot intelligence increases at the deeper levels of the tree, but so does the time it takes to figure out the best move.
Awarded the Graduate Pathway Scholarship (2019) by Illinois Institute of Technology covering 25% tuition fee of Masters Education.
Volunteered at an NGO named Disha for their ‘Speak Initiative’. This aimed to develop soft-skills among less-privileged kids of a charity-run school in India. It was observed that the students had better communication skills by the end of this Initiative. They were able to convey their ideas better, had improved class participation and were overall more confident in their approach.