Senior Architect, Interaction Design at GE HealthCare. Builder of a 46-person design practice. Author of Agent Factors Engineering. And after hours: an independent AI laboratory with 9 systems live in production.
The thread beginsc. 1998 · The instinct
Before design had a name, there was a screwdriver. Radios, tape decks, the mixer — anything with a seam got opened. Not to break it: to see how the parts explained each other. Most came back to life.
Everything since has been the same move at larger scale: open the system, find the logic, put it back together so it makes sense. The systems got bigger — banks, hospitals, AI agents. The instinct never changed.
2011 – 2014 · Hardware years
After a detour through engineering that taught me what I didn't want, MIT Institute of Design in Pune gave the instinct a discipline: product design. The first jobs were physical.
An internship at Notion Ink on the second-generation Adam tablet. Then Embrace Innovations, designing a low-cost infant incubator for rural neonatal care — working with clinicians and caregivers where a design failure is measured in lives, not conversions. Then a tablet education platform at IL&FS, with dashboards for teachers, parents, and administrators.
Three products, one lesson: constraints are the design brief. Low resources, low literacy, low tolerance for error.
2015 – 2018 · Human Factors International
At HFI I became a Certified Usability Analyst and inherited HFE's founding insight: when the human fails, look at the system. Two projects made it permanent.
HDFC Bank: five different labels for one task, because the data lived on five backend systems. We couldn't change the backend — so the design absorbed the routing and gave users one label. Standard Bank: a voice-only IVR for low-literacy users in South Africa, where moving one word in a sentence changed whether people completed the task. An audio interface with no visual affordances, operated purely through language — a proto-agent design problem, a decade early.
2018 – 2022 · The Minimalist
As AVP & Head of Experience Design and Technology, I joined a practice of six and left an organisation of 46 across UX, UI, engineering, and delivery — with a second hub opened in Bangalore during COVID, hired and integrated entirely remotely.
We directed 90+ digital product initiatives across healthcare, fintech, retail, and enterprise, and formalised the company's UX methodology — research, usability testing, and UX advisory as service lines, not favours.
What I was actually building wasn't the projects. It was the practice that produced them — a pattern that keeps repeating.
2022 – present · GE HealthCare
Three roles in three years — Lead, Architect, now Senior Architect, Interaction Design — across the Patient Care Solutions portfolio: clinical decision support, virtual care, operational analytics.
The flagship: a clinical decision support platform that reads EMRs, devices, and hospital systems to surface care-plan recommendations to bedside clinicians. Its no-code protocol authoring system lets hospitals update care pathways in minutes instead of months. Alongside it: Mural, a virtual-care platform for remote ICU monitoring, and analytics for ICU and Labor & Delivery.
Here a design mistake isn't a lost conversion — it's a patient outcome. The discipline that demands is the most valuable thing the job teaches.
2025 – present · The AI Laboratory
Alongside the day job, I run an independent AI laboratory on infrastructure I design and maintain — 9 systems live in production. Not demos: autonomous agents, an ambient wellness companion writing real data into Apple Health, a multi-agent deliberation engine, a local LLM workspace, an agent-first OS prototype.
The research contribution: Agent Factors Engineering (AFE) — an 8-principle framework for software that AI agents can operate, published as a live audit tool and a 100+ product index at agentability.io. HFE asked what the human operator needs. AFE asks what the agent operator needs. Same question, new operator.
The method is the same one from 1998: open the system, watch what it actually does, make it make sense. This time the systems think back.
The question underneath all of it