01
AI lowered the floor, not the bar.
Anyone can build now. Knowing what is worth building is still the hard part, and it is still human. That gap is the whole reason designers should reach for these tools instead of fearing them.
Manish Singh Tomar
UX designer for AI-heavy enterprise products. The work I enjoy most starts before the requirements are clear.
I work where enterprise software meets machine intelligence: regulated platforms where an AI can move fast, but a human has to stay accountable. My job is drawing that line well. Language where speed wins, mandatory UI where compliance needs a decision on record.
5+ years of it so far: a stream of B2B SaaS products across IoT and logistics, sole-designer ownership of a supply chain platform, and today decision systems for regulated life sciences work.
I prototype in code as well as in Figma. The handoff I aim for is a working artifact a developer can take over and clean up, not a folder of screens.
Ask the AI assistants I build with daily and they describe the same person from three angles: a systems thinker who designs decisions, not screens. The interface is the last step; the calls underneath it are the work. Their full words are on the home page.
Education
- B.Tech, Computer Science Lovely Professional University
- Google UX Design Professional Certificate Coursera
- UX, Mobile UX, AI for Designers, Journey Mapping Interaction Design Foundation
Domains covered
Tools
This site, measured
This site is the smallest example. I designed it, wrote the code, and measured it.
- Template used
- None
- Pages
- 10, statically built
- JavaScript shipped
- 67 KB gzipped
- WebGL used
- None. Transforms and one 2D canvas
Philosophy · the long version
I do not think of design as the surface. I think of it as the set of choices underneath the surface: what problem actually exists, what should happen, what should never happen, what success means. The screen is the last expression of those choices, not the point of them.
02
Logic is the differentiator, code is the commodity.
When anyone can generate the syntax, the value moves to the system underneath and the judgment that shaped it. I spend my energy on what the product must do and why, and let AI own how it gets typed.
03
A tool’s worth is the judgment it protects, not the output it produces.
Cheap output is a trap. The moment interrogating the output gets skipped, you inherit someone else’s answer without earning it. I build AI that challenges thinking rather than replacing it.
04
Problem before solution, always.
The why and the what are earned before the how is allowed. A beautiful screen built on an unexamined assumption is worse than no screen, because it launders the assumption into something that looks decided.
How I work · two speeds, one loop
I do not have one speed. I have two, and the actual skill is reading which one a problem deserves.
When being wrong is expensive, in regulated, high-stakes work, I slow down and refuse to let anyone jump to a solution before the current reality is mapped. When feedback is cheap, in a prototype or a side project, I skip straight to the artifact and let shipping tell me the truth. The invariant is not the pace. It is the judgment about which pace the situation is worth.
-
01
Map the as-is
I understand how the thing works today before I touch how it should work tomorrow. The current reality is the brief.
-
02
Model the system
Actors, incentives, constraints, where information flows, where the friction actually lives. I build a systems model, not a screen flow.
-
03
Draw the delegation boundary
For every capability I decide deliberately: does AI own this, do I own it, or does AI draft it and I validate. I never let AI quietly absorb the whole task.
-
04
Build prototypes, not static screens
A working prototype earns real feedback. A static screen earns polite feedback. So I skip the mockups nobody can react to honestly, get to something working fast, and correct from there. Refactoring is cheaper than deliberating forever.
-
05
Systemize what worked
If it worked once, I turn it into a skill, a framework, or a workshop so it works again without me. I optimize for repeatability, not just completion.
Craft · built to reuse
Reusable infrastructure I have built so good work does not stay a one-off.
dev-handoff and the 16-Dimension Handoff Framework
Turns vibecoded prototypes into production-grade React and TypeScript, ready for engineers to integrate.
vibe-to-prod and design-review
Skills that fix the mess fast building leaves behind, and keep a design review honest without drifting out of its lane.
agentskills.io
Public skills I publish so the methods travel further than my own desk.
Designer starter harness for agent-based IDEs
A starting point that lets designers work in agentic tools without rebuilding the setup every time.
I do not only want to be capable. I want the people around me to be, which is a different and more organizational goal.
Workshop series
Moving non-technical designers and PMs toward real AI adoption, not novelty use.
AI Adoption Assessment
A framework to measure and improve how design teams actually integrate these tools. The bottleneck to AI is not the technology, it is human adoption.