npm i react-ghost-text
React, TypeScript, CSS
React component to show ghost text suggestions in an input field (line in VS Code or Gmail Smart Compose)
I am a PhD student in Computer Science at Cornell University advised by Aditya Vashistha. Previously, I was a Research Fellow at Microsoft Research and a Student Researcher at Google DeepMind.
Research Interests
My work sits at the intersection of Natural Language Processing and Human-AI Interaction, where I evaluate and mitigate cultural and social biases in large language models. My aim is to make generative systems value-aligned and globally safe. My research has been published at top-tier venues, open-sourced by Microsoft, and covered by mainstream media. On the side, I love tinkering with AI and building things!
Background
Before Cornell, I graduated magna cum laude in Computer Science from Ashoka University, India. My undergraduate research earned me the Undergraduate Research Excellence award, while my campus leadership was recognized with the Student Life Excellence award. Beyond academics, I've competed at varsity-level in badminton and football/soccer, and represented at the national level in tennis.
Resources for Students/Applicants
- Practical strategies for PhD applications based on my experience.
- Advice for prospective Quad Fellowship applicants.
Awards
- LinkedIn PhD Fellowship, 2025
- Quad Fellowship, 2023
- Cornell University Fellowship, 2022
- Methods Recognition Award, CSCW 2021
- Best Paper, COMPASS 2020
- Undergraduate Research Excellence Award (Top 2)
- Student Life Excellence (<1%; for social impact)
Publications
1. Fluent but Culturally Distant: Can Regional Training Teach Cultural Understanding?
Dhruv Agarwal, Anya Shukla, Sunayana Sitaram, Aditya Vashistha
Under Review New
2. AI Suggestions Homogenize Writing Toward Western Styles and Diminish Cultural Nuances
Dhruv Agarwal, Mor Naaman, Aditya Vashistha
CHI 2025 Press Coverage
arXiv | The Atlantic | Fast Company
3. Steering AI-driven Personalization of Scientific Text for General Audiences
Taewook Kim, Dhruv Agarwal, Jordan Ackerman, Manaswi Saha
CSCW 2025 New
4. Let's Talk it Out: Mitigating Social Bias in LLMs using Agentic Dialogue
Dhruv Agarwal
In progress
5. "One Style Does Not Regulate All": Moderation Practices in Public and Private WhatsApp Groups
Farhana Shahid, Dhruv Agarwal, Aditya Vashistha
CSCW 2025
Projects for Fun
Ashoka Bot
Python (+ Flask), Facebook APIs
A Messenger bot I developed as a weekend hack to give Ashoka University users information about mess menu, shuttle timings and important phone numbers. The bot has received and sent over 80k messages to over 600 unique users since it was started in March 2017.
VAHN: Vehicle Accident Hazard Notifier
TF Object Detection API, Socket.io, Python, Node.js
A real-time smart city system for automatic detection and reporting of road accidents, using a machine learning and vision algorithm, for reduced relief response time and avoiding chaining of accidents. Adjudged 5th best project out of 400 teams at HackIIIT-D 2018.
Energy Sensing System
Arduino, Raspberry Pi, Python (+ Flask), MySQL
We developed a device that could be placed in dorm rooms at Ashoka to monitor wastage of energy. We used multiple sensors to detect whether energy was being wasted in the form of unnecessary use of lights and AC, and displayed the information on a user-friendly web app.
Beating Facial Image Obfuscation
Keras, Python, SciPy stack
A machine learning model (autoencoder) to beat common forms of facial obfuscation like blurring and pixelation. In addition, we defined another obfucation model, "blocking", to beat our model and hence show limitations of our model. We also tried generative models like GANs and variational autoencoders, which did not give good results for this task.
Thor: Online Code Judge
Python (+ Flask), PHP, MySQL
An online code judge (like CodeChef) intended to be used for programming contests within my University. This platform was eventually used by the instructor for programming assignment submissions in the CS101 course in Spring 2016.