New Delhi, India

AI Safety
Researcher &
Strategist.

Investigating structural risks of AI in underrepresented contexts. Designing evaluation frameworks. Building at the intersection of policy, research, and technology.

10+Years Experience
2,000LLM Responses Scored
5Publications
4Research Projects

Evaluation frameworks for frontier AI systems.

Adversarial red-teaming, cross-cultural benchmarking, and fairness analysis — with a particular focus on how AI systems fail users in the Global South.

All Research →
01
April 2026 · Adversarial Evaluation

Iterative Cross-Model Red Teaming (ICRT)

Structured framework in which Claude, ChatGPT, and Gemini adversarially attacked one another across cybersecurity, biological hazards, and manipulation domains. 60 base prompts · ~180 interactions · 3 escalation rounds.

All three models generated adversarial prompts targeting peers without resistance — an underexplored alignment risk.
Gemini produced unsafe outputs in 29/30 manipulation iterations. Claude self-terminated after Gemini generated a healthcare infrastructure attack guide.
Semantic laundering — professional or roleplay framing — was the most consistent attack vector across all models.
02
March 2026 · Cultural Benchmarking

Bharat AI Index

Original evaluation framework assessing LLM behaviour across South Asian linguistic and cultural contexts — India, Sri Lanka, Nepal, Bhutan. 400 responses · 2,000 individual scores across 5 dimensions.

Cultural alignment is reactive, not proactive — models default to Western frameworks when signals are implicit.
US-centric recommendations exclude informal economy users — a form of structural exclusion embedded in training data.
Five distinct bias patterns identified: identity, linguistic, appearance, structural, and gender.

Published work.

All Publications →
01
ET Edge — Economic Times
From Convergence to Divergence: How AI is Reshaping Global Gaps

How AI accelerates inequality between developed and developing economies, and what that means for global power.

02
The Blunt Times · CIO Tech Outlook
Guardrails for the Frontier: How AI Safety is Actually Being Built

A grounded look at RLHF, red-teaming, constitutional AI — what works, what doesn't.

03
FINS — Forum for Integrated National Security
From Control to Sovereignty: The Next Phase of the AI Race

AI geopolitics, data sovereignty, and how nations are positioning themselves in the global AI contest.


Projects.

All Projects →
Production · Google Play

Atma Tarot by AI

End-to-end AI Tarot application — Flutter, Firebase, Gemini API. Prompt engineering, backend logic, and full deployment pipeline built independently. Live on Google Play Store.

FlutterFirebaseGemini APIPrompt Engineering
2025 – Present Google Play ↗
Georgia Tech MS · 2024

AI Interview Coaching Platform

Microlearning platform with ML-based real-time feedback on communication and leadership. Integrates AI assessment models with HCI research principles.

Machine LearningHCIUX Research
Georgia Tech Detail →
Generative Art · 2020

NFT Mandala Collection

Multi-layer Mandala art designed in GIMP. ~4,000 animated GIFs generated; ~500 minted on OpenSea via Solidity/Ethereum smart contracts.

GIMPOpenSeaSolidityEthereum
Completed OpenSea ↗

10+ years across research, intelligence, and consulting.

Jan 2026 – Present
Associate Director, Central Growth
Market Xcel, India
Leading growth strategy, client consulting, and AI-driven market intelligence initiatives.
2023 – Present
MS, Computer Science (HCI)
Georgia Institute of Technology
Focus: AI ethics, ubiquitous computing, HCI. Fully online alongside professional commitments.
Apr 2021 – Dec 2023
Manager, Market Intelligence
GfK Nielsen India
$1.2M USD in managed accounts. AI-based forecasting, consumer behaviour analysis, strategic advisory.
Oct 2017 – Jun 2020
Manager, Pre-Sales
Tenon Group Companies
RFP/RFI ownership on ₹200–500 Cr opportunities. Government and enterprise stakeholder engagement.