Aman Khullar

Hi, nice to have you here! I am a CS PhD student at Georgia Tech, Atlanta Campus, where I am working with the TanDEM lab, under the inspiring and caring guidance of Dr. Neha Kumar.

Through my work I study Global Development through AI. Currently I am thinking about visiblizing work of ASHAs, remote caregiving, and evaluation of LLMs in public health. I feel at home when I am building systems and am currently learning how to conduct qualitative and ethnographic studies.

I have previously worked with Dr. Michael L. Best at Georgia Tech and have spent time in industry with the Alexa ASR team at Amazon and the Siri TTS team at Apple.

Before this, I worked as a Research Associate at Gram Vaani (GV), under the inspiring guidance of Dr. Aaditeshwar Seth. At GV, I tried to answer how to appropriately design and deploy AI (mostly speech) systems in a social context to empower the under-represented people using an Interactive Voice Response (IVR) platform.

Email  /  CV  /  Google Scholar  /  GitHub  /  LinkedIn

In my research, I take a human-centered approach towards global development. Adopting Dr. Amartya Sen and Dr. Jean Drèze's argument of achieving freedom as a means and ends, I study how AI can support the well-being of individuals. I also think about power structures and how technologists have an important role to play in engaging with the topogoraphies of power as argued in this book. Please write to me if you are interested in similar research areas or if you need help in getting started (though I am a fledgling myself!), or if you are in Atlanta and have some nice bookstore recommendations.

In this work, we propose a data augmentation approach that addresses the problem of lack of data for online hate speech detection in limited data contexts using synthetic data generation techniques.

ACM Journal on Computing and Sustainable Societies, Jan 2024

In this work, through an iterative design process and detailed user feedback, we describe several nuances in running voice-based surveys, and compare their accuracy of data collection through equivalent keypress-based surveys.

ACM COMPASS, June 2021

In this study, we present our findings in terms of the time and cost-savings made through the introduction of these tools, and describe the feedback of the moderators towards the acceptability of AI-based automation in their workflow.

India HCI, Nov 2021 and KDD Workshop in Data Science for Social Good, July 2021

In this study, we report on the feasibility of running automated question-answering systems in the context of rural and less-literate users in India, accessed through IVR (Interactive Voice Response) systems.

ACM COMPASS, June 2021

This study provides an evolving portrait of the health, economic & social consequences of the COVID-19 pandemic on vulnerable populations in India.

Health Promotion International, August 2020 (published May 2021)

This paper presents a sequence-to-sequence trimodal hierarchical attention-based model which pays more attention to the text modality and analyzes the use of audio in deep multimodal abstractive summarization architectures.

NLPBT Workshop, EMNLP 2020

This paper presents an amalgamation of a token classifier and a sentence classifier for sentence boundary disambiguation in noisy PDF documents.

FinNLP Workshop, IJCAI-PRICAI 2020

Through my previous experience and projects, I have been able to work with a highly interdisciplinary team at Gram Vaani to understand the actual needs of the users and we modified our product scope accordingly (even if that meant starting from the ground-up). I also keep going back to lessons learnt under Aadi Sir's guidance by visiting this webpage regularly. Even at Product Labs, I observed the challenges faced in taking a product from ideation to production and then incorporating the feedback from the users. The projects I describe next are also a result of significant team effort and collaboration.

The proposed model, MMBiDAF is used for for carrying out the task of multimodal summarization which has been inspired from the various previous state of the art models existing in the literature. This model was developed since it encompasses all the input modalities, calculates the similarity between them and then uses a multimodal attention layer on top of image-aware and audio-aware texts to get an output distribution over the source document.

This is my PyTorch implementation of the paper, 'Deep Knowledge Tracing' by Chris Pieth et al. I took up the re-implementaiton of this paper in order gain a greater insight in the field of computer supported education while further improving my implementation skills of building end-to-end deep learning models.

We designed and fabricated an autonomous rover for Mars Rover Challenge. This project was a combined applied research effort by student from branches encompassing computer science, mechanical, electronics and biotechnology. Our 26 member team together built the autonomous Mars Rover and were the only college in North India to reach the top 71 teams in world. Following year we qualified among top 30 teams to compete in Poland for the European Rover Challenge.

The model combines the idea of Bi-Directional Attention Flow (BiDAF) network along with high performing bidirectional Gated Recurrent Units.

The purpose of this project was three fold:

1. Description and analysis of basic as well as advanced solution concepts of game theory.
2. Theoretical explanation and practical implementation of machine learning algorithms on open source datasets.
3. Interaction of the two fields and their application in state of the art research.

I implemented this algorithm from scratch in C++ to explore the field of combinatorial optimization.

I believe that we can be the change that we wish to bring in this world! I am working in the field of computer science in order to apply it to help in solving the existing problems of the world. With the widespread availability of internet, the field of computer science can really exert its influence for social good.
Various organisations are focusing their work on the application of technology for Social Good and that is something that excites me and makes me believe in the potential of technology when used in the right manner.
Moroever, I love interacting with people and listening to their disrupting ideas about how we all can work together to make this world a better place! If you share a similar philosophy or just want to contact me about my work, I would love to hear from you!

This website has been inspired by Jon Barron.