AI Product Designer + Builder

I design complex AI products people can understand, trust and control.

Senior Product Designer working across AI, SaaS and enterprise products — from product strategy and UX to working prototypes and launch-ready websites.

Senior Product Designer across AI, SaaS and enterprise products — from strategy to prototypes and launch-ready websites.

12+ yrs product design

AI + enterprise SaaS

Strategy → Design → Build

View selected work

Start a project

AI Decision ConsoleAwaiting approval
01
ObjectiveCut false-positive alerts on the overnight shift
Set
02
PlanRe-rank alerts, group by camera, draft shift summary
4 steps
03
ExecutionRe-ranking and grouping applied across the shift
Running
04
Confidence0.00 — below the 0.85 threshold for auto-apply
0.00
05
Human approvalA reviewer must approve before the summary is sent
Awaiting
06
CompletionBlocked until the approval is recorded
Blocked
Every action is logged and reversible

AI Decision ConsoleAwaiting approval
01
ObjectiveCut false-positive alerts on the overnight shift
Set
02
PlanRe-rank alerts, group by camera, draft shift summary
4 steps
03
ExecutionRe-ranking and grouping applied across the shift
Running
04
Confidence0.00 — below the 0.85 threshold for auto-apply
0.00
05
Human approvalA reviewer must approve before the summary is sent
Awaiting
06
CompletionBlocked until the approval is recorded
Blocked
Every action is logged and reversible

01 — Selected work

Selected work where the real design problem was understanding the system behind the screen.

A system knows something. A person has to decide what to do about it.

Enterprise Computer Vision Platform

Enterprise AI / Computer Vision

Enterprise AI / Computer Vision

Product Design

Product Design

When AI detects something wrong, what does a human need to see before acting on it?

The platform reads camera feeds and flags events that need attention. Detection was never the hard part — the work was deciding what an operator sees, and in what order, before they act on something the system raised.

01

Source

02

Detection

03

Alert

04

Evidence

05

Human review

06

Decision

07

Record

Computer vision — system view

Schematic

A · Alerts

Camera 04

Restricted Zone

Needs review

Camera 11

Zone B

Acknowledged

Camera 02

Zone C

Acknowledged

Context stays attached to the detection

B · Evidence

0.87

Camera 04

Restricted Zone

14:32

Confidence 0.87

C · Review

Confirm incident

Dismiss

B · Evidence

0.87

Camera 04

Restricted Zone

14:32

Confidence 0.87

C · Review

Confirm incident

Dismiss

Evidence is surfaced before action

D · Record

Acknowledged · recorded 14:33

Actions remain traceable

Evidence is surfaced before action

Confidence communicates uncertainty

Context stays attached to the detection

Human review remains part of the decision

Actions remain traceable

Interface intentionally abstracted due to NDA.

Evidence is surfaced before action

Context stays attached to the detection

Actions remain traceable

Read the case study →

Enterprise Network Topology Platform

Telecom / Network Operations

Telecom / Network Operations

Product Design / UX Architecture

Product Design / UX Architecture

When a network is too complex to see at once, how do you help someone find the one thing that needs attention without losing its context?

Operators are responsible for more elements than anyone can hold in mind at once. The design work was making narrowing feel like navigation — so that reaching one node never means losing sight of what it connects to.

Operators are responsible for more elements than anyone can hold in mind at once. The design work was making narrowing feel like navigation — so that reaching one node never means losing sight of what it connects to.

01 · Estate

01 · Estate

All elements

All elements

No filters applied

No filters applied

02 · Segment

02 · Segment

Selected segment

Selected segment

Region: North

Region: North

SEGMENT ASEGMENT BSEGMENT CSEGMENT DSEGMENT ESEGMENT F
SEGMENT ASEGMENT BSEGMENT CSEGMENT DSEGMENT ESEGMENT F
SEGMENT ASEGMENT BSEGMENT CSEGMENT D

03 · Cluster

03 · Cluster

Affected cluster

Affected cluster

Region: North

Region: North

Tier: Core

Tier: Core

Node 041Node 042Node 043Node 044Node 045Node 046Node 047Node 048Node 049Node 050Node 051Node 052Node 053Node 054Node 055Node 056Node 057Node 058Node 059Node 060Node 061Node 062Node 063Node 064
Node 041Node 042Node 043Node 044Node 045Node 046Node 047Node 048Node 049Node 050Node 051Node 052Node 053Node 054Node 055Node 056Node 057Node 058
Node 041Node 042Node 043Node 044Node 045Node 046

04 · Node + dependencies

04 · Node + dependencies

Focal node

Focal node

Region: North

Region: North

Tier: Core

Tier: Core

State: Degraded

State: Degraded

Node 012

Node 018

Node 067

Node 068

Node 071

Node 074

Node 041

Upstream

Node 012

Node 018

Focal

Node 041

Downstream

Node 067

Node 068

Node 071

Node 074

Highlighted paths show what a change here would reach

Interface intentionally abstracted due to NDA.

02 — Services

Where I usually come in.

Where I usually come in.

I work with AI and SaaS teams at three points — understanding a difficult system, designing how it should work, and building it far enough to test or launch.

I work with AI and SaaS teams at three points — understanding a difficult system, designing how it should work, and building it far enough to test or launch.

01

Product UX

Product UX

For products where the difficulty is in the system, not the screen. Workflows, roles, states, permissions and decisions have to become clear before an interface can help anyone.

For products where the difficulty is in the system, not the screen. Workflows, roles, states, permissions and decisions have to become clear before an interface can help anyone.

Typical work

Discovery · UX architecture · Flows, roles and states · Product UI · Design systems · AI interaction design

Discovery · UX architecture · Flows, roles and states · Product UI · Design systems · AI interaction design

Best when

the product does something genuinely complicated and people can’t tell what it’s doing or what to do next.

the product does something genuinely complicated and people can’t tell what it’s doing or what to do next.

02

AI / SaaS Website + Framer

AI / SaaS Website + Framer

For teams whose product is stronger than their explanation of it. I learn how the product actually works first, then design and build the site around that understanding — not the other way round.

For teams whose product is stronger than their explanation of it. I learn how the product actually works first, then design and build the site around that understanding — not the other way round.

Typical work

Product understanding → Positioning and story → Information architecture → Visual design → Responsive Framer build → Launch

Product understanding → Positioning and story → Information architecture → Visual design → Responsive Framer build → Launch

Best when

the product makes sense when you explain it yourself, but the website doesn’t yet do that job for you.

the product makes sense when you explain it yourself, but the website doesn’t yet do that job for you.

See how this site was designed and built →

03

Prototype Sprint

Prototype Sprint

For an idea, a workflow or an AI concept that has to exist before anyone can judge it. A short engagement that ends with something running in a browser rather than a deck.

For an idea, a workflow or an AI concept that has to exist before anyone can judge it. A short engagement that ends with something running in a browser rather than a deck.

Typical work

Product thinking · Interaction design · Framer · AI-assisted development · Working browser prototypes

Product thinking · Interaction design · Framer · AI-assisted development · Working browser prototypes

Best when

you need to see the idea working before you can make the next product decision.

you need to see the idea working before you can make the next product decision.

Have something in mind? Start a project →

03 — How I work

I don’t stop at the screen.

I don’t stop at the screen.

I start underneath the interface — understanding the problem, structuring the complexity, designing the interaction, and making ideas real enough to test and learn from.

01 —

Understand

What is actually happening?

Users

Business context

Constraints

Existing workflows

01 —

Understand

What is actually happening?

Users

Business context

Constraints

Existing workflows

02 —

Structure

What needs to become clear?

Information architecture

Flows

Roles

States

Relationships

02 —

Structure

What needs to become clear?

Information architecture

Flows

Roles

States

Relationships

03 —

Design

What should people see, decide and control?

Product UX

Interaction

UI

Design systems

Prototypes

03 —

Design

What should people see, decide and control?

Product UX

Interaction

UI

Design systems

Prototypes

04 —

Build

How quickly can we make the idea real?

Working prototypes

Framer

AI-assisted development

Browser-based experimentation

04 —

Build

How quickly can we make the idea real?

Working prototypes

Framer

AI-assisted development

Browser-based experimentation

The tools change. The job doesn’t: make complexity understandable enough to act on.

04 — Lab

Things I’m learning by building.

Things I’m learning by building.

Experiments around AI agents, interaction design and new ways of taking ideas from thought to working software.

Experiments around AI agents, interaction design and new ways of taking ideas from thought to working software.

Experiment 01

What should happen when an AI agent needs a human decision?

What should happen when an AI agent needs a human decision?

An exploration of how people supervise autonomous agents — from goals and actions to confidence, approvals, failures and audit trails.

An exploration of how people supervise autonomous agents — from goals and actions to confidence, approvals, failures and audit trails.

AI Decision Console · Run 04

AI Decision Console

Experiment

Agent run

Objective

Summarise last night’s incidents for the operations handover

Plan

01

Collect overnight events

Done

02

Group by source

Done

03

Draft handover summary

Done

04

Send to operations

Blocked

Action / tool use

collect_events

Complete

group_by_source

Applied

send_summary

Awaiting approval

Human decision

Confidence

0.72

0.72 · threshold 0.85

Below the threshold for acting without a person

Decision

Approve

Ask for changes

Reject

Activity

14:31

objective set

14:31

plan generated

14:32

tools called

14:33

confidence below threshold

14:33

paused for a person

Send remains blocked until a human decides

Experiment 02

Process experiment

Process experiment

Can a designer direct agents instead of handing off screens?

Can a designer direct agents instead of handing off screens?

This site is part of the experiment. I set the design direction and rules first, then use AI agents to turn them into working UI. My role shifts from producing every screen myself to making the decisions, reviewing what comes back and refining it.

This site is part of the experiment. I set the design direction and rules first, then use AI agents to turn them into working UI. My role shifts from producing every screen myself to making the decisions, reviewing what comes back and refining it.

The result is the site you’re using now — designed, built and made responsive in Framer.

Design direction

CLAUDE.md

Claude Code

Framer

Framer Agent

Human design review

05 — About

I’m Karunakar. I like the part before the interface becomes obvious.

I’m Karunakar. I like the part before the interface becomes obvious.

I’ve spent more than 12 years designing complex digital products across enterprise, SaaS and AI.

I’ve spent more than 12 years designing complex digital products across enterprise, SaaS and AI.

Much of my work has involved systems where the interface is only the visible layer — network operations, computer vision, analytics, workflows, roles, alerts and decisions underneath.

Much of my work has involved systems where the interface is only the visible layer — network operations, computer vision, analytics, workflows, roles, alerts and decisions underneath.

Today I’m increasingly interested in what happens when software can act on behalf of people, and what product designers need to design when the user is no longer the only one taking actions.

Today I’m increasingly interested in what happens when software can act on behalf of people, and what product designers need to design when the user is no longer the only one taking actions.

Karunakar Kadari

12+ years in product design

Enterprise · SaaS · AI

I work from Hyderabad, India, and collaborate with teams building difficult products.

I work from Hyderabad, India, and collaborate with teams building difficult products.

06 — Start a project

Tell me what you’re working on.

If you’re working on an AI product, a SaaS platform, a website or an early idea that’s hard to explain or hard to use, send me a little context — I’ll tell you whether I can help.

No pitch needed. A few sentences is enough.

No pitch needed. A few sentences is enough.

Or reach me directly

karunakar.kadari6@gmail.com

LinkedIn ↗

I’ll use the details you share to respond to your enquiry.

I’ll use the details you share to respond to your enquiry.

Create a free website with Framer, the website builder loved by startups, designers and agencies.