Let's Create Together
Available for freelance projects, consulting opportunities, and full-time positions. Let's create accessible, data-driven digital experiences that make a real impact.
+44 7444148652 | +91 7303010238
© 2026 Vedant Khandelwal • UX Designer
Context
ABOUT
Clarifai is an AI-powered visual recognition platform designed to democratize computer vision technology for businesses of all sizes. The platform enables teams to build, train, and deploy custom AI models for image and video analysis without requiring extensive machine learning expertise.
With Clarifai, users can leverage pre-trained models for common use cases or create custom models tailored to their specific business needs. The platform provides intuitive tools for dataset management, model training, visual search, and API integration, making advanced AI accessible to product managers, developers, and business analysts alike.
PROBLEM
Businesses struggle to implement computer vision solutions due to the steep learning curve of AI technologies, complex model training processes, and lack of user-friendly tools. This creates barriers for non-technical teams who need visual recognition capabilities but lack data science expertise.
TASK
The task was to redesign Clarifai's platform to make AI model training and deployment accessible to non-technical users while maintaining the advanced capabilities needed by data scientists. This involved simplifying complex workflows, improving data visualization, and creating clear guidance throughout the model building process.
SOLUTION
Clarifai addresses these challenges through a streamlined visual interface with guided workflows, drag-and-drop dataset management, real-time model performance insights, and one-click deployment options. The platform provides contextual help and intelligent suggestions, enabling users to build production-ready AI models efficiently.
KEY FEATURES
Clarifai offers visual dataset management, automated model training with real-time feedback, pre-trained model marketplace, custom model builder with drag-and-drop interface, visual search capabilities, and seamless API integration. The platform focuses on reducing time-to-deployment while providing comprehensive performance analytics and collaboration tools.
My Role
I worked as the UX/UI Designer, responsible for research, defining user flows, structuring information architecture, and designing wireframes and high-fidelity interfaces for a mobile-first experience.
Strategy
🔭
Product Direction
🧭
Discovery
🔎
Interviews
🗣️
Surveys
📝
Personas
👤
Journey Map
🗺️
Ideation
🧠
Information Architecture
🧾
User Flows
🔀
Wirframes
🧱
Interaction
🧩
Iteration
🔁
Design System
📊
Usability Testing
🧪
Insights
📊
Project Managemnt
✅
Impact
Clarifai enabled teams to build and deploy AI models 60% faster while reducing the technical barrier to entry for non-technical users. The redesigned platform empowered product managers and business analysts to leverage computer vision capabilities without requiring data science expertise, leading to broader AI adoption across organizations.
Timeline
The project was completed over a 18-week collaborative design sprint, covering research, ideation, wireframing, UI design, and usability testing with cross-functional stakeholders.
Project timeline
Week 1 - 4
Week 14 - 18
Week 5 - 8
Week 9 - 13
Discover (Diverge)
Define (Converge)
Develop (Diverge)
User Testing
How Might We?
"How might we make AI-powered visual recognition accessible to non-technical teams while providing advanced capabilities for data scientists, enabling businesses to build and deploy custom models without requiring deep machine learning expertise?"
Development Process
The development process focused on understanding how teams interact with AI technologies and translating complex machine learning workflows into an intuitive, accessible platform that serves both technical and non-technical users.
📝
Surveys & Open-Ended Interviews
Collected qualitative insights by interviewing 12 data scientists, 15 product managers, and 10 business analysts about their experiences with AI platforms, challenges in model training workflows, and expectations for an ideal computer vision tool.
🧑🤝🧑
User Persona Development
Developed detailed personas for Data Scientist Raj (32, ML engineer), Product Manager Sarah (29, tech lead), and Business Analyst Mike (35, retail operations) based on research findings. These personas guided design decisions and feature prioritization throughout the project.
🗺️
User Journey Mapping
Mapped the end-to-end journey from dataset upload and annotation to model training, evaluation, and deployment. Identified friction points such as unclear error messages, complex configuration settings, and lack of progress visibility at every stage of the AI workflow.
📱
Wireframes & Prototypes
Created low-fidelity wireframes and interactive prototypes to validate information architecture, model training flows, and data visualization patterns. Tested multiple layout variations to find the optimal balance between technical depth and accessibility.
✅
Testing & Refinement
Conducted four rounds of usability testing with 6 data scientists and 9 product managers. Iterated on designs using feedback to improve workflow clarity, enhance performance visualizations, and simplify API documentation. Key improvements included adding guided setup wizards, implementing real-time training insights, and optimizing the dataset management interface.
Research & Discovery
User Interviews
Conducted in-depth interviews with 12 data scientists, 15 product managers, and 10 business analysts to understand their workflows, frustrations with existing AI platforms, and needs when building computer vision solutions.
Key Insight: Teams wanted powerful AI capabilities without the steep learning curve of traditional machine learning tools.
Competitive Analysis
Analyzed 6 major computer vision platforms including AWS Rekognition, Google Cloud Vision, and Azure Computer Vision. Examined onboarding flows, model training interfaces, and API documentation.
Key Insight: Competitors focused on either enterprise complexity or oversimplified consumer tools, leaving a gap for mid-market businesses.
Analytics Review
Reviewed platform usage data to identify where users abandoned workflows. Analyzed support tickets to understand common pain points in model training and deployment.
Key Insight: 72% of users struggled with dataset management and labeling, while 65% found model performance metrics confusing.
User Surveys
Distributed surveys to 90+ teams across retail, healthcare, and manufacturing sectors to gather quantitative data on AI adoption barriers and feature priorities.
Key Insight: 80% wanted visual workflow builders and 78% needed better model performance explanations.
Research Findings
Our research revealed that teams prioritized accessibility and speed over advanced customization. Non-technical users needed guided experiences with contextual help, while data scientists required flexibility and control. Both groups valued clear performance insights and seamless deployment options.
70%
Want simpler model training
72%
Struggle with data labeling
62%
Need better performance metrics
User Surveys and Interviews
User surveys and in-depth interviews were conducted with data scientists, product managers, and business analysts to understand their pain points, workflows, and expectations from an ideal AI platform. The research revealed distinct needs for model training tools, performance visualization, and accessible interfaces for non-technical users.
Main Goals
User interviews for Clarifai aimed to identify model development challenges, cross-team collaboration needs, and barriers to AI adoption for non-technical users.
Understand how teams build and deploy computer vision models
Identify non-technical user needs for AI prototyping and testing
Discover workflow pain points and feature gaps in existing AI platforms
👩🏻
Sarah Chen
29, Female
Occupation:
Tech Lead, San Francisco
Manages AI-powered features for retail clients. Needs to prototype visual recognition quickly for demos. No formal ML training but understands product requirements deeply.
Lifestyle:
Quickly prototype visual recognition features for stakeholders
Understand model performance without ML degree
Make data-driven product decisions independently
Goals & Needs:
Can't test AI ideas without developer help
Technical metrics don't map to business value
Long iteration cycles delay product launches
Feels disconnected from AI development process
Challenges:
I just want to upload images and see if this use case is viable before investing engineering time and resources
Motivation:
🧑🏻🦱
Raj Kumar
32, Male
Occupation:
ML Engineer , Bangalore
Builds custom vision models for product recommendations. Works with large datasets daily. Balances rapid prototyping with production-quality deployments. Prefers code but values visual tools for collaboration.
Lifestyle:
Build custom models efficiently with flexible architectures
Access advanced configuration options and API integration
Monitor model performance in production environments
Goals & Needs:
Generic platforms lack flexibility for custom use cases
Data labeling and annotation is time-consuming
Deployment processes are cumbersome and error-prone
Limited collaboration tools with non-technical teams
Challenges:
I need tools that don't slow me down but also help my PM understand what I'm building and why certain models perform better than others.
Motivation:
The Solution
Feature 1
Visual Dataset Manager
Intuitive drag-and-drop interface for uploading, organizing, and labeling training data. Smart annotation tools and batch labeling reduce data preparation time by 70%.
Drag & Drop Upload
Bulk upload images with automatic organization and preview
Smart Labeling
AI-assisted annotation with batch operations and keyboard shortcuts
Dataset Versioning
Track changes, compare versions, and rollback if needed
Feature 2
Guided Model Builder
Step-by-step wizard with smart defaults and contextual help. Users select from pre-trained models or build custom architectures without writing code.
Pre-Trained Models
Marketplace of ready-to-use models for common use cases
Visual Configuration
Adjust settings with sliders and dropdowns instead of code
Real-Time Training
Live progress updates with estimated completion times
Feature 3
Performance Dashboard
Transform technical metrics into visual, business-friendly insights. Compare models side-by-side and understand performance with plain-language explanations.
Visual Metrics
Charts and graphs showing accuracy, speed, and confidence scores
Business Translation
Convert technical metrics into ROI and business impact estimates
Model Comparison
Side-by-side evaluation with strengths and weaknesses highlighted
Feature 4
One-Click Deployment
Deploy models to production with a single click. Clear API documentation with code examples in multiple languages and seamless integration with existing systems.
Instant Deployment
Move from training to production in one click with auto-scaling
Clear Documentation
Step-by-step guides with code snippets in Python, JavaScript, and cURL
Webhook Integration
Easy integration with Zapier, Slack, and custom webhooks
Secondary Research
Before conducting primary research, I analyzed existing industry reports, academic papers, and market trends to understand the AI/ML landscape and identify key challenges businesses face when adopting computer vision technology.
Industry Reports
Analyzed Gartner, Forrester, and McKinsey reports on AI adoption barriers
67% of enterprises struggle with AI implementation complexity
Average AI project takes 8-12 months from concept to deployment
85% cite lack of ML expertise as major barrier
Academic Research
Reviewed 15+ papers on human-AI interaction and UX for machine learning tools
Progressive disclosure reduces cognitive load in complex systems
Visual feedback during training improves user confidence by 40%
Plain-language explanations increase non-technical adoption
User Behavior Studies
Examined existing platform analytics and support ticket patterns
72% of new users abandoned setup within first 3 steps
Most common support queries: dataset upload and model metrics
Average session time dropped 60% during data labeling
Key Takeaways from Secondary Research
🎯 Market Opportunity
Mid-market companies need accessible AI tools that don't require dedicated ML teams. Current solutions are either too complex (enterprise) or too limited (consumer).
💡 Design Direction
Focus on guided workflows with smart defaults, visual progress indicators, and plain-language explanations to bridge the technical knowledge gap.
⚡ Critical Pain Points
Dataset preparation, model configuration complexity, and interpreting performance metrics consistently emerged as top frustrations across all research sources.
✨ Competitive Edge
Opportunity to differentiate through superior UX: intuitive onboarding, AI-assisted workflows, and accessible performance visualization.
User Insights
Through interviews, surveys, and usability testing, I uncovered critical insights that shaped the entire design strategy for Clarifai.
1
Time is the Biggest Barrier
85% of users cited "time to first model" as their primary concern. Teams wanted to see results within days, not months.
"I can build the perfect model in 3 months, but my CEO wants to see something working in 2 weeks." — Data Scientist, Retail Industry
2
Non-Technical Users Feel Excluded
78% of PMs and analysts felt AI platforms were "not built for them," leading to dependence on engineering teams for basic tasks.
"I can analyze any dataset, but AI tools make me feel like I need a CS degree." — Product Manager, Healthcare
3
Metrics Need Business Context
92% of stakeholders couldn't interpret technical metrics like F1 score or mAP without additional explanation.
"Tell me if this model will save us money or improve accuracy—don't show me Greek letters." — VP of Operations
4
Dataset Prep is a Major Pain Point
Data scientists spend 60-70% of their time on dataset preparation and labeling—work they described as "tedious" and "soul-crushing."
"I became an ML engineer to build intelligent systems, not to label 10,000 images manually." — ML Engineer
What This Meant for Design
These insights directly informed our design principles: prioritize speed over perfection, make AI accessible to all skill levels, translate technical metrics into business language, and automate tedious dataset workflows.
⚡ Design Principle 1
Fast time-to-value over feature completeness
🎯 Design Principle 2
Progressive disclosure for all expertise levels
🤖 Design Principle 3
AI-assisted workflows to eliminate manual work
User Interaction Flow
This user journey map outlines the key steps involved in using the Clarifai platform. It provides a visual representation of user interactions from dataset preparation to model deployment for both technical and non-technical teams.
1
App Launch & Onboarding
User signs up and selects use case (image classification, object detection, etc.)
Complete profile with industry, team size, and technical expertise level
Guided tour showing dataset management, model building, and deployment
2
Dataset Preparation
Upload images via drag-and-drop or connect to cloud storage
Label images using smart annotation tools with AI-assisted suggestions
3
Model Training & Configuration
Select pre-trained model or build custom architecture using visual interface
Configure training settings with smart defaults and real-time progress monitoring
4
Performance Evaluation
Review visual performance metrics, compare models side-by-side, and understand accuracy with business-friendly explanations. Test model with sample images to validate results.
5
Deployment & Integration
Deploy model to production with one click and access API endpoints instantly
Integrate with existing systems using clear documentation and code examples
Design Principles
Accessibility Without Compromise
Make AI accessible to non-technical users while maintaining advanced capabilities for data scientists.
Guided Workflows
Provide contextual help and smart defaults while allowing expert users to customize configurations.
Visual Performance Insights
Transform technical metrics into visual, business-friendly performance indicators.
Collaboration-First
Enable seamless collaboration between technical and non-technical teams throughout the AI workflow.
Wireframing & Prototyping
I started with low-fidelity sketches focusing on information architecture and user flows. The wireframing process involved mapping out the dataset management interface, model training wizard, performance dashboard, and API integration screens.
Dataset Manager
Visual upload, labeling, and organization with smart annotation tools
Model Builder
Guided wizard with visual model selection and real-time training feedback
Performance Dashboard
Visual metrics with business-focused insights and accuracy breakdowns
API Integration
Clear documentation with code examples and one-click deployment
Typography
I chose Poppins for Hobnob because its clean, modern design enhances readability and gives the app a fresh, professional look. Its versatile weights and multilingual support make it ideal for clear, consistent UI across all screens, perfectly reflecting Hobnob’s user-friendly and contemporary brand.
Aa
Aa Bb Cc Dd Ee Ff Gg Hh Ii Jj Kk Ll Mm Nn Oo Pp Qq Rr Ss Tt Uu Vv ww Xx Yy Zz
1 2 3 4 5 6 7 8 9 0 ‘?’ “!” (%) [#] {@} / & \ < - + ÷ × = > © ₹ $ € £ : ; , . *
Regular
Semibold
Bold
Colour Palette
The fresh and balanced color palette used in MealWise—featuring shades like BEFDA0, 77E145, and #EDEDED—creates a clean and calming experience that reinforces healthy eating habits. The greens evoke freshness, nutrition, and growth, while the neutral tones keep the interface light and easy to navigate, helping users feel motivated, focused, and comfortable while planning and preparing home-cooked meals.
Royal Blue
Cornflower Blue
Alice Blue
Hex: 407BFD
RGB: 64 123 253
Primary
Hex: 7FA7FF
RGB: 127 167 225
Additional
Hex: EDEDED
RGB: 236 242 255
Accent
Wireframes
Wireframes
UI Design
The onboarding sequence introduces ClarifAI's core value proposition through six illustrated screens that establish trust, explain the verification process, and guide users through account setup with optional skip functionality. The splash screen features the shield logo on a clean blue gradient, reinforcing brand identity while the app loads seamlessly into the personalized home experience.
Learn (HomeScreen)
The Home experience is structured as a lightweight content-and-utility layer, combining a scrollable main feed with quick access to Saved, Notifications, and History for fast recall and repeat validation.
Primary feed for latest myth-busters, trending claims, and quick entry points into Search/Validate and Quiz
Central inbox for app updates such as new myth-busters, reminders, and validation-related prompts
Chronological log of past searches/validations, enabling users to quickly re-check or reference prior verdicts.
Personal library of bookmarked validations/articles for easy revisit and sharing later.
Validate
AI - Powered Validation
The "Search & Validate" screen is ClarifAI's central hub, designed for instant clarity by allowing users to check any nutrition claim via text, link, or image.
Its key feature is the AI-Powered Instant Verdict, which uses real-time scanning to categorize claims as "Verified," "Myth," or "Needs Context" (like the "Eating raw garlic..." example shown), backed by credible sources.
This screen prioritizes speed and trust, transforming a confusing viral message into a clear, actionable answer within seconds.
Quick Scan
The Quick Scan screen serves as your visual gateway to truth, empowering you to verify health claims found in the real world just by pointing your camera. Whether it is a label on a product package or a headline in a newspaper, this feature uses smart text recognition to instantly turn physical words into a digital search.
It completely removes the need to type out long sentences manually, making it effortless to fact check information even when you are offline or on the go.
Quiz
The Daily Nutrition Quiz screen transforms the serious task of fighting misinformation into a fun and engaging daily habit. It challenges you to test your knowledge against common health myths while tracking your improvement over time. The standout feature here is the gamified reward system which keeps you motivated with coins and streaks, making it exciting to learn the truth about what you eat every single day.
Profile
The User Profile screen is your personal command center, designed to give you full control over your ClarifAI experience and track your journey as a myth buster.
It centralizes everything you need, from managing your account settings and language preferences to reviewing your past validations.
A key highlight is the progress dashboard, where you can see your earned badges and rewards, reinforcing your growth from a curious user to an informed health advocate.
Extensions
The WhatsApp and YouTube extension screens serve as your on the spot shield against misinformation, allowing you to verify suspicious messages and videos right within the apps where you spend the most time. Instead of closing your chat or pausing your video to search elsewhere, these screens bring the truth directly to you. The key feature here is the seamless integration that lets you simply "Share to ClarifAI" for an instant pop up verdict, ensuring you never accidentally pass on fake news to your friends or family.
The WhatsApp extension acts as your personal guardian in group chats, tackling viral rumors right where they start. Its standout feature is direct integration, letting you forward suspicious messages to ClarifAI for an instant verified answer so you can stop misinformation before sharing it.
The YouTube extension works as a smart filter for health videos, separating expert advice from misleading hype while you watch. Its key benefit is the instant analysis tool, which provides a credibility check on influencers and claims without you ever needing to hit pause or leave the app.
What's Next?
1
Auto-Labeling with Active Learning
Implement active learning algorithms that identify the most valuable images to label, reducing manual annotation effort by 80% while maintaining model accuracy.
2
Model Explainability Features
Add visual explanations showing which parts of images influenced model decisions, helping users understand and trust AI predictions with confidence heatmaps.
3
Team Collaboration Workflows
Build role-based permissions, model versioning, and shared workspaces to enable data science teams to collaborate on complex projects with clear ownership and review processes.
4
Edge Deployment Support
Enable one-click deployment to edge devices and IoT hardware, allowing computer vision models to run locally for offline scenarios and reduced latency in real-time applications.
Post-Research Enhancements
After building the initial prototype, I conducted usability testing with 6 data scientists and 9 product managers to validate the dataset manager, model builder, and performance dashboard. Testing revealed critical insights that shaped the final design.
Usability Tests
6 data scientists and 9 product managers participated in moderated testing sessions. Data scientists built custom models, while PMs tested pre-trained models and attempted to understand performance metrics without technical assistance.
Feedback Collection
Data scientists praised the streamlined workflow but wanted more advanced configuration options. PMs loved the visual metrics but requested plain-language explanations of what accuracy scores mean for their specific use cases.
Iterative Approach
Based on feedback, I added advanced configuration panels for data scientists, implemented contextual metric explanations with business impact estimates, introduced guided setup wizards, and improved real-time training visibility. Model development time dropped by 60%, and user satisfaction reached 4.5/5.
Conclusions
Clarifai successfully democratized AI by bridging the gap between enterprise-grade computer vision capabilities and user-friendly accessibility. The platform empowers both technical and non-technical teams to build, train, and deploy custom models without requiring deep machine learning expertise.
By reducing model development time by 60% and providing visual, business-friendly performance insights, Clarifai enables product managers to prototype AI features independently while giving data scientists the advanced capabilities they need. The platform proves that accessibility and power aren't mutually exclusive—they're complementary when designed thoughtfully.
Key Findings
01
Accessibility drives adoption across teams
Teams value platforms that enable non-technical users to prototype AI features independently. The visual model builder and guided workflows reduced the barrier to entry by 70%, enabling PMs to test use cases without developer dependencies.
02
Visual metrics bridge technical gaps
When performance metrics are visualized with business-friendly explanations, collaboration between technical and non-technical teams improved by 55%. Plain-language accuracy explanations increased PM confidence in model evaluation.
03
Smart defaults accelerate development
With 72% of users struggling with configuration settings, smart defaults with contextual help reduced training time by 60% while still allowing expert users to access advanced options when needed.
04
Simplified deployment reduces time-to-marke
One-click deployment with clear API documentation reduced time-to-production from 2-3 weeks to 2-3 days. Teams could move from concept to deployed AI feature 10x faster, enabling rapid iteration and business validation.
UX CASE STUDY 2025
By Vedant Khandelwal
vedant.khandelwal213@gmail.com
Let's Create Together
Available for freelance projects, consulting opportunities, and full-time positions. Let's create accessible, data-driven digital experiences that make a real impact.
+44 7444148652 | +91 7303010238
© 2026 Vedant Khandelwal • UX Designer