Industry research

Research I’ve led inside product organizations: large-scale studies of how people work with AI, turned into frameworks, design guidance, and tools that product teams adopted.

Microsoft · Tech Futures (X3)

Applied Science Intern

Jun – Aug 2026 · Redmond, WA
Mentors: Alice Ferng, Cinoo Lee, Peter Jahn
1,534 knowledge workers surveyed22 interviews80 design patterns21 human-factors challenges448 developers surveyed
  • Led a large-scale study involving 1,556 knowledge workers (1,534 surveyed, 22 interviewed) on what makes a great co-worker in an AI-native workplace. Built the BACI framework of co-worker qualities that apply to both humans and AI, identified 11 co-worker archetypes, and a taxonomy of AI work etiquette.
  • Turned years of human-centered AI research into the Design Cookbook for UI/UX of AI: 80 patterns across 10 design dimensions, mapped to 21 human-factors challenges.
  • Led a study of 448 developers on where and why they draw the line on AI autonomy, with implications for agentic tooling and meaningful work.
  • Co-led an investigation into agent-to-agent coordination in large multi-agent workplaces.
Co-worker preprint →Design Cookbook →AI autonomy preprint →Slides →
Microsoft Research · SAINTES

Ph.D. Research Intern

Jun – Sep 2025 · Redmond, WA
Mentors: Christian Bird, Robert DeLine, Carmen Badea
860 developers surveyed8 causal experiments with 2,000 developers22 AI systems specified2 Distinguished Papers
  • Led a mixed-methods study of 860 developers on where AI delivers real value in software engineering. Built a cognitive task-appraisal framework and specified 22 AI systems developers want built, with the Responsible AI guardrails needed to ship them (ICSE-SEIP 2026; IEEE Distinguished Paper, ICSME 2026).
  • Ran eight experiments with 2,000 developers on transparency patterns and cognitive-forcing interventions. Targeted friction improved verification and reduced over-reliance while preserving the experience.
  • Co-designed a matched-pair study of individual and team AI use and identified the Productivity Pressure Paradox (ACM SIGSOFT Distinguished Paper, ICSE 2026).
  • Translated findings into product practice with AI for Healthcare, AI Ethics, and Copilot Bing Search, delivering guidance and evaluation rubrics the teams adopted.
Dashboard →ICSE-SEIP paper →ICSME paper →ICSE paper →Slides →
MOSIP · IIIT Bangalore

Global Impact Research Fellow

Mar – Jun 2025 · Gates Foundation-funded fellowship
89% F1 scoreValidated on academic courseware & industry products7 productsMulti-agent LLM workflow
  • Designed an end-to-end multi-agent workflow that automatically debugs usability issues in national ID systems, turning inclusivity decision rules into AI-driven checks.
  • Built an extensible framework that adapts to diverse personas and inclusivity dimensions (cognitive, socio-economic, gender), with recommendations to improve equity in national identity platforms.
  • Each screen evaluation has the resource cost (water and CO₂) of about 2–4 Google searches, 3–5× less than chatting with an LLM directly.
Apply-<x>Mag paper →Apply-<x>Mag tool & artifact →