Unlocking English fluency for 300 million Indians — not through lessons, but through the confidence to speak.
They don't need grammar lessons. They need the confidence to speak — and we built the engine that unlocks it.
They don't lack knowledge. They lack the confidence to speak. And every product in the market is solving the wrong problem.
Optimize for streaks and DAU, not learning outcomes. Users feel busy but can't hold a conversation.
One dimension of a multidimensional problem. Speaking ≠ confidence. Accent training doesn't prepare you for a job interview.
₹200-500/session doesn't scale. Quality varies by tutor. Session 100 isn't smarter than session 1.
Every metric moved through systematic experimentation — not luck. And every metric maps directly to a user who stayed longer, engaged deeper, and renewed.
Users who engage more during their trial period renew at dramatically higher rates:
4+ day active users grew from <10% (pre-Jun'25) → 20% (Jan'26). 7-day active: <2% → 5%. Both directly proportional to M0 renewal growth.
Revenue held steady at ~₹5 Cr/mo for 6 months while we rebuilt the retention engine. M0 doubled. D1 tripled.
The next phase scales acquisition behind the engine that now works.
After studying thousands of users across India, we identified the real barrier — and designed the only system that addresses it.
Users know English grammar. But the moment they need to speak, shame takes over. Every mistake confirms their worst fear: "I'm not good enough."
English learning competes with Instagram, YouTube, and WhatsApp for the same 30 minutes. Content must be emotionally engaging, not just educationally correct.
Users see "English speakers" as a different kind of person. Until English feels like part of their identity — not a skill they're acquiring — they won't stick.
AI companions in story worlds where mistakes are safe, practice feels like play, and confidence builds through emotional experience — not grammar drills.
Every user gets a deeply personalized experience. Here's what it looks like in practice.
Composite archetype
from
1,000+ user data points
25, Indore. BBA graduate. Works at her father's garment business. Dreams of an MBA abroad but freezes in English conversations.
Wants to confidently handle international business calls and prepare for IELTS — without anyone knowing she's "practicing English."
Placed in a startup storyline — AI characters include her mentor (supportive), a skeptical investor (challenging), and a co-founder (peer).
Persona-matched users like Riya renew at 2x the rate of generic users. Engagement time is 40% higher.
That emotional recognition — the moment a user thinks "this was made for someone like me" — is why persona-matched users renew at 2x the rate of generic users.
Three layers. One intelligence. Every user interaction makes the product smarter. Every internal decision makes the next user's experience better. The loop never stops.
Every experiment, decision, pattern, and user signal. When a team member leaves, their knowledge doesn't.
Built-in decay detection. When behavior shifts, the system flags its own knowledge as stale.
Actively relearns. Hypotheses from 6 months ago are tested against current reality.
When one subsystem discovers a pattern, that insight automatically informs every other subsystem.
This loop runs continuously. Every cycle sharpens all three layers — making the next cycle even faster.
90.6% CM1 proves the model. YTD EBITDA positive at ₹7.42 Cr. Last quarter, we deliberately reinvested — doubling spend toward higher-quality users while M0 doubled. The engine works. Now we scale it.
H1: EBITDA positive at conservative spend. H2: Deliberately reinvested — doubled ad spend toward higher-quality users. Result: M0 doubled, retention cracked.
₹112 Cr in reserves. We're raising from strength — capital was preserved while we solved retention. Now that the engine works, this capital deploys behind proven economics.
Every user makes the persona mapping sharper. Every sharper persona makes the content better. Every better content generates richer signals. The system doesn't just run — it accelerates. And the longer it runs, the wider the lead.
240 archetypes, refined through 180K+ real users. Each new user makes the mapping sharper for the next.
12+ months of AI conversation iteration. Every session teaches the system what resonates and what falls flat.
Thousands of real interactions across India's diverse contexts. This can't be built in a lab — only in the field.
Optimized for streaks and DAU. Treats India as one segment. No persona intelligence, no cultural depth.
Strong on speaking mechanics. But pronunciation ≠ confidence. One dimension of a multidimensional problem.
Real conversations, but ₹200-500/session economics. No compounding — session 100 is identical to session 1.
The retention, the economics, the personalization — all stem from this
compounding system.
Every month it runs, the
product gets smarter, the economics get stronger, and the lead gets
harder to close.
12 months. Two phases. First, deepen the system until we understand every user archetype at a level no competitor can reach. Then, scale — because every new user will instantly feel "this was made for me."
Build the moat so wide that no competitor can cross it.
When every new user instantly feels "this was made for me."
Unit economics positive throughout. EBITDA returns as retention gains compound at scale.
A lean team augmented by an AI first approach that gives us the output of a company 10x our size.
Built SpeakX from zero to profitability.
Architect of the AI infrastructure powering the three-layer intelligence system.
Designed persona intelligence framework & product-org learning loops to drive personalisation.
We're raising a Series B to capture a $5.9B market with the only product that has cracked retention, unit economics, and compounding intelligence — all at once. ₹112 Cr in reserves. This is a raise from strength, not necessity.
Accelerate the evaluate → predict → self-evolve loop. More persona depth, faster classification, richer signals per cycle.
Voice AI, real-time classification, content generation at near-zero marginal cost. 10x users without 10x the cost.
$5.9B market, zero competition that works. 4x LTV/CAC on matured cohorts. Persona-matched funnels ready to scale.
90.6% CM1. LTV/CAC positive across all CAC levels. Deploy capital behind what already works.
Others teach English. We built the intelligence system that
learns
why each person fails — and rewires itself to unlock them.
