Aatman Vaidya

prof_pic.jpg

aatmanvaidya [at] gmail.com

I am a SDE/ Researcher at Tattle, where I contribute to building citizen centric open-source tools and datasets to understand and respond to online harms and abuse in India.

I work at the intersection of language, technology and society. My research interests broadly lie in computational social science and human centered AI, with a focus in measuring and mitigating harms like hate speech and misinformation particularly in low-resource contexts and making digital spaces safer and more inclusive.

Here are some research directions I work on:

  • Understanding spread of information on online platforms and designing strategies to mitigate it. I particularly study how harmful content propagates on social networks like Twitter, Koo and Gab. I also examine the cross-platform spread of misinformation, like how health-related misinformation and financial scams spread across platforms (ongoing work).
  • Community-led participatory approach for online safety. I strive to build safer and culturally well rounded AI systems that are centered on lived experiences of people at the end of online violence.. Through expert-led participatory approaches, I aim to understand the context-specific ways that harm manifests in different regions and communities. This approach has informed our development of AI safety benchmark dataset on sex and hate-related crimes in Hindi and dataset for online gender-based violence in India.
  • Building user centric tools to combat online abuse. At Tattle, I help build tools like Uli, a browser plugin that redacts slurs and abusive content from your online feed; Feluda, an engine for analysing multi-lingual and multi-modal content; Deepfakes Analysis Unit, a collaborative space for analyzing deepfakes and AI generated content and Viral Spiral, a multiplayer card game about sharing news on the internet.

You can find more information about me from my CV.

updates

May 2025 Presented at the Tech4Dev AI Cohort Program on incorporating Safety in AI for good applications.
May 2025 Got featured by FOSS United as a part of Maintainers Month where they feature 31 Indian FOSS projects in the month. (see Folklore page)
Mar 2025 Gave a lightning talk on Open Source Deepfake Detection at MisinfoCon India 2025. (slides)
Mar 2025 Gave a talk on “Evaluating Indian Language Performance in LLMs” at the AI for Global Development by Agency Fund. (slides)
Jan 2025 Conducted a suvery of Indic Language Capabilities in LLMs. (preprint)
Nov 2024 Worked with MLCommons to develop an AI safety bechmark dataset in Hindi on Hate and Sex-realted crimes in India.
Apr 2024 The Uli Dataset paper is accepted at Workshop on Online Harms and Abuse at NAACL 2024. (Won the Outstanding Paper Award!!)
Dec 2023 Conducted a Shared Task on Gendered Abuse Detection in Indic Languages at ICON 2023.
Nov 2023 Our Paper titled, “Analysing the Spread of Toxicity on Twitter” got accepted at CODS-COMAD.
Nov 2023 Invited for a pannel at Digital Citizen Summit.
Sep 2023 Our Paper titled, “Forecasting the Spread of Toxicity on Twitter” got accepted at IEEE CogMI
Aug 2023 Started working at Tattle !
May 2023 Went to Cluj, Romania to participate in the semi-final round of the Bosch Future Mobility Challenge - Built an autonomous driving car