Himanshu Kalra · Bangalore → Berlin
Four-plus years of research across cybersecurity, fintech, entertainment, fashion and online media, in consultancy and embedded. Currently at Trellix, studying how people decide to trust AI when the stakes are high.
Four-plus years of research, most of it in cybersecurity. Enterprise, regulated, B2B. Currently at Trellix, studying how security analysts decide to trust an AI verdict when the stakes are high, and where they stop.
Four-plus years of research on complex professional tools, and on expert users who already know how to do the job without them. Currently at Trellix, working on enterprise security software. Adoption, defaults, and what a redesign quietly takes away.
Four-plus years of research on high-stakes decisions under regulation. Cybersecurity now, retail investing before it. Currently at Trellix, studying when people accept an automated verdict and when they override it. Earlier, user requirements for a stocks-investment app at Navi.
Four-plus years of research, and the habit of building the thing rather than only specifying it. I've built 117 German worksheets my partner and I use daily, an image pipeline for a catalogue of 62 objects, and GenAI tooling that halved study turnaround at a consultancy. Currently at Trellix, studying how people decide to trust AI when the stakes are high.
Answer · Reversibility is not reliability
OutcomeReframed the product's autonomy architecture before launch, and reset three shipping decisions about what an AI may act on, how it is corrected, and how its confidence is shown.
Answer · The most important column wasn't on the list
OutcomeCut 27 candidate columns to 8 evidence-backed defaults, surfaced a missing priority-zero field before build, and reset the hash and timeline decisions.
Answer · The redesign experts rejected
OutcomeDiagnosed the adoption regression across a multi-round program of 45 participants in 11 organizations, plus a 67-analyst operations study. Prioritized three Tier-1 blockers and drove fixes into the roadmap. Two of four shipped within a release.
UX Researcher III in cybersecurity. B2B, enterprise, regulated. Security analysts get an AI verdict on every alert. Some believe it, some override it, and the difference isn't what the vendors think. Three case studies from this work are above.
117 interactive worksheets my partner and I use to learn German for a life abroad. They review what we keep getting wrong, and sync progress across devices.
An editorial catalogue of 62 objects that have outlived their trends. A custom image pipeline re-shot every photograph into a uniform studio look. Made with Sharmi. Launching soon.
Associate, then Senior UX Researcher at a strategic foresight and innovation consultancy. Led and supported nine-plus generative and evaluative studies for clients including Google and YouTube. Led the team's GenAI integration. Mapped our research processes, then built custom tools that cut study turnaround in half while improving quality.
User requirements research for the MVP of a stocks-investment app. Two hundred-plus one-on-one user interactions to uncover friction across funnels, surveys to size and prioritise the issues, moderated usability testing to close the loop.
Field research with drivers, warehouse managers and contractors for an employee-focused project. Interviews and analysis surfaced their needs and pain points, and streamlined the internal system.
Business and editorial strategist at a media platform for architects and designers. Coordinated a hundred-plus writers across eight-plus countries, picked topics through keyword research, analytics and reader surveys, and co-facilitated a workshop on architectural journalism.
Registered Architect in practice in Jaipur. Co-authored and presented case studies on citizen-initiated sustainable projects. The research habit started here too: fifty-plus interviews about eating habits for a food-delivery startup, as a design intern in Ahmedabad.
Two studies published in full, with the corpus, codebook, method and stated limitations. The findings can be checked rather than taken on trust.
76 coded items from 372 Hacker News threads, 2021–2026: what technical people do when a security warning gives them no answer they can act on. Ships the corpus, the sampling log, the codebook, and a disclosed-limitations section.
The quantitative companion. 10,042 public security-incident records, and what the human role in them looks like once you count.
Plus the tools the studies run on, at github.com/uxrhimanshu.
Next step
I'm a UX Researcher III at Trellix, relocating to Germany. I'm looking for a senior research role on a team that already knows why it wants one. I need sponsorship. My Master's is anabin-recognised, so the EU Blue Card route is open.
60 days notice · Bangalore, UTC+5:30 · happy to talk in your evening