<!--
Work info
-->
Role:
Lead Product Designer
Timeline:
Jun 2023 - Apr 2024
Scope:
UI Design, UX Research and GTM collaboration
<!--
Context
-->
Turning APIs into something customers could try
NeuralSpace had strong speech APIs, but evaluating them required code. VoiceAI made that evaluation possible through a simple web platform where teams could test STT/TTS outputs first, then integrate with confidence.
As the sole product designer, I took VoiceAI from early definition to MVP launch and iteration. I shaped the workflows, MVP scope, and usability loop, working daily with engineering and aligning with sales/marketing on demo readiness.
Outcomes
500+
active users within 24 hours of MVP launch
Usability
positive feedback from customers and sales teams
+8
enterprise customers onboarded
90% of language AI ignores Arabic and Indian languages.
90% of language AI solutions are built for European languages, while billions of people speak Arabic and Indian languages. VoiceAI was part of NeuralSpace’s push to make speech AI usable for these regions.
This demo shows how VoiceAI lets teams evaluate Arabic STT/TTS first, then integrate.
<!--
Process
-->
How the first version took shape
I aligned with the CEO, CTO, product lead, and engineering lead to understand what VoiceAI needed to ship and the constraints we were working within. I also brought sales and marketing in from day one, since they would be demoing and positioning the product later.
What came out of these conversations
VoiceAI needed to support evaluation before API integration
The experience had to be step-by-step and jargon-free
We should ship a focused MVP first, then expand based on feedback
GTM alignment early would keep the product and messaging consistent
<!--
MVP
-->
Lo-fi: getting the structure right before visuals
I started with working in lo-fi to make sure the product felt straightforward. Most iterations happened on homepage and output screen, because those two screens defined whether the product felt clear or overwhelming.
Lo-fi decisions
Made the main workflows large, visible entry points on the homepage (no hunting through navigation)
Kept transcription configurations easy to find, but secondary
Placed API integration after the workflow, once users had results to evaluate
Hi-fi: aligining with rebrand and shipping the MVP
After lo-fi, I moved into high-fidelity using NeuralSpace’s rebrand and design system. The focus was to keep the UI minimal and consistent, while making output-heavy screens easy to scan during demos and real use.
The initial MVP was intentionally focused. We shipped the smallest set of workflows needed to validate core transcription usage and demand.
MVP scope
Real-time transcription
File-based transcription
Sentiment analysis + transcript summaries
Translation with side-by-side comparison
<!--
Testing
-->
Testing: validating the MVP
After the MVP launch, we ran task-based user testing to see where people hesitated, got stuck, or lost confidence. We set objectives upfront so findings mapped to concrete fixes, not opinions.
<!--
Post MVP
-->
We fixed the moments that made users doubt.
A faster way to get value from transcripts
GPT-style interactions were becoming familiar, so we introduced AMA as an optional layer to help users get answers quickly without digging through transcription.
No more guessing during uploads
Testing showed a repeat pause during long uploads. People couldn’t tell if VoiceAI was working. We added clear progress states so the system always communicates what’s happening.
Making speaker views easy to switch
Speaker separation was previously tucked inside configurations. Testing showed it was too hidden for something so central to reading transcripts. We made it default and added a dropdown on the output page to switch views.
Beyond transcription: adding Text-to-Speech
We expanded VoiceAI beyond transcripts by introducing Text-to-Speech. This supported NeuralSpace’s focus on Arabic and Indian languages, where pronunciation and dialect accuracy matter.
Together, these changes shifted VoiceAI from a transcription tool into a flexible environment for exploring speech-based workflows.
<!--
Reflection
-->
Beyond the Interface: Designing the Go-To-Market
To bring VoiceAI into the market, I partnered with marketing to translate the product into campaign assets that were easy to skim and easy to share.
Learnings and Takeaways
This project reinforced my ability to:
Take a technically complex AI system and turn it into workflows people can actually understand and use
Design for both technical and non-technical users without building two separate products
Make product decisions that balance usability, system constraints, and go-to-market needs
Work closely with engineering, product, data science, sales, and marketing to ship and position a real product
Core takeaway: I can design AI products that are grounded in how systems work, but still feel clear, usable, and ready for real-world adoption.
