Building DuoAccent in public: our journey from idea to beta
I learned my American accent at 15, playing Call of Duty with Americans on voice chat. I'm from Bangalore. Nobody taught me. Nobody coached me. I just spent hundreds of hours talking to Americans in-game, and my brain started mimicking the sounds.
At the time, I didn't think much of it. It was just gaming. But years later, when I started my lead gen agency selling to US companies, I realized that accent was the single biggest reason my close rates were 3x higher than other Indian salespeople I knew. Same scripts. Same products. Same market. The only difference was that when I got on a call, Americans didn't hear “offshore.” They heard “one of us.”
I built a multi-five-figure MRR agency on the back of that advantage. And every time I hired a new rep who struggled with accent bias, I thought the same thing: I had an unfair advantage. Most offshore sales reps don't. What if we could give it to everyone?
That question became DuoAccent.
Why existing solutions fall short
Before building anything, I spent weeks trying every accent and pronunciation tool on the market. Here's what I found:
BoldVoice
Good product, great marketing. But it's built for general English learners, not sales professionals. The practice content is generic sentences, not cold call scripts. And there's no conversational practice -- you're always repeating phrases, never having a real back-and-forth.
ELSA Speak
Impressive AI, but also aimed at general English learners. Strong on pronunciation scoring, weak on sales-specific training. No roleplay, no gamification that actually drives daily habit formation, no focus on the specific sounds that trigger bias in business contexts.
Accent coaches
The gold standard for quality, but at $100-200/hour, totally inaccessible for most offshore reps. A BPO with 200 agents can't hire 200 accent coaches. And even individually, the scheduling and cost make it unsustainable for daily practice.
The gap was clear: nobody was building accent training specifically for sales professionals, with sales-specific content, gamification that drives daily habits, and AI-powered conversational practice. That's a very specific product for a very specific audience. And it didn't exist.
The tech decisions
When you're building a product that needs to ship fast and iterate faster, every technical decision matters. Here's what we chose and why.
React Native + Expo
Cross-platform from day one. Our users are on both iOS and Android, and we can't afford to maintain two native codebases this early. Expo's managed workflow lets us ship fast without fighting with native build tools.
Next.js
Server-rendered, fast, great SEO. The landing page and blog need to rank on Google. Next.js makes that easy with server components and built-in metadata handling.
OpenAI Whisper API
Best-in-class speech-to-text. We need the transcription to be accurate across heavily accented speech -- and Whisper handles this better than any alternative we tested. The word-level timestamps are critical for our scoring model.
Custom scoring model
This is where our real IP lives. We built a custom model that takes Whisper's output plus raw audio features and produces scores across Clarity, Pronunciation, Rhythm, and Confidence. We also detect specific phoneme substitutions and map them to bias impact scores.
Firebase + Dodo Payments
Firebase for auth, database, and storage. Dodo Payments for subscriptions and billing. Both let us move fast without building infrastructure from scratch.
The 20-day sprint
We set ourselves a constraint: build and launch the beta in 20 days. Not because we're masochists, but because constraints force decisions. When you have unlimited time, you build unlimited features. When you have 20 days, you build what matters.
Here's roughly how those 20 days broke down:
Landing page, waitlist, blog. Get the story out there, start collecting emails.
Instant Accent Demo. The viral hook. Record, transcribe, score, display results. This was the hardest part -- getting the scoring model to produce meaningful, consistent results.
Mobile app core: onboarding, daily drills (all 4), scoring integration, streak tracking, XP system. The gamification layer that makes people come back every day.
Auth flow, user profiles, progress tracking, leaderboard. Everything needed to make the app feel complete and social.
Polish, bug fixes, beta invite flow. Ship it.
What we cut: the Sales Simulator (too complex for v1), achievement badges (nice-to-have), social features beyond the leaderboard, and accent comparison (seeing how you compare to the average speaker from your language background). All of these are on the roadmap. None of them were essential for validating the core idea.
The beta promise
If you're joining the beta, here's what you're getting:
Free access to the core daily drills during beta
Founding member pricing -- locked in at the lowest rate we'll ever offer, forever
Direct access to me -- I'm personally onboarding every beta user and reading every piece of feedback
Early access to the Sales Simulator when it launches
In return, I need honest feedback. What works. What doesn't. What feels broken. What you wish existed. Beta users aren't just early adopters -- they're co-builders.
The long-term vision
Here's where this gets big. Every time someone uses DuoAccent, we collect a data point: a sample of accented speech, paired with our scoring output. Over time, we're building the world's largest dataset of accented sales speech.
That dataset is the moat. With enough data, we can train a proprietary accent scoring model that's dramatically more accurate than anything built on generic speech data. We can detect not just which sounds are wrong, but predict which corrections will have the highest impact on sales outcomes. We can build accent-specific coaching that's personalized down to the individual user, their native language, their target accent, and their specific weak patterns.
The vision: DuoAccent becomes the world's most accurate accent scoring engine for professional speech, powered by the largest dataset of accented business communication ever assembled.
That's the endgame. But right now, the goal is simpler: help 1,000 offshore sales reps close more deals by removing accent as a barrier. One daily session at a time.
Why I'm building in public
I could have built this quietly and launched with a polished product page. But I believe in building in public for the same reason I believe in the product itself: transparency builds trust.
You'll see our metrics. Our struggles. Our wins. The features that work and the ones that don't. I'll share the revenue numbers when there are revenue numbers. I'll share the user feedback, good and bad.
If you're an offshore sales rep who's ever lost a deal because of how you sound rather than what you know, this product is for you. And I want you to be part of building it.
Join us
The waitlist is open. Beta invites go out in waves. Every waitlist member gets founding member pricing and priority access to new features.
If you're interested in following the build, I post updates regularly. And if you have ideas, complaints, or just want to chat about accent bias in sales -- my DMs are open.
Join the waitlist
Get free beta access, founding member pricing, and be part of building the future of accent training for sales professionals.
Join the waitlistWritten by Basheer Ahmed, Founder of DuoAccent