Stimuler

The Highest-Rated Mobile App for English
Speaking Practice. Globally.

Google Play's Best AI App 2023
Winner - Consumer Mobile
AWS Generative AI Accelerator 2025
Only consumer mobile app globally selected
02 · The Team

Built by college friends who have been in this space for 6+ years

Akshay Akash
Akshay Akash
CEO & Growth

World Championship debater. IELTS 8.5 Band. Built the MATS distribution system from scratch. Leads all growth and marketing.

Ankit Pandey
Ankit Pandey
AI

Previous experience in ML and Linguistics research. Leads the speech understanding and assessment AI teams.

Anesh Srivastav
Anesh Srivastav
Product

Consulting experience with international governments. Product supply stint at Procter & Gamble. Leads product strategy, design, and user experience.

Akshat Baranwal
Akshat Baranwal
Development

Built and shipped multiple mobile apps since college. Leads all engineering, infrastructure, and app development.

03 · Traction

Where We Are Today

01
12M+
Downloads
Global installs across iOS and Android
02
$3.8M
ARR
Annual Recurring Revenue as of 15 May 2026
03
~4.4x
YOY Growth
Revenue growth over the prior year
04 · Market

1.5B people are learning English
none of them solved for fluency at scale

$40B
Global language learning market
~$2B
Spent every year on mobile apps
DUOLINGO
$1000M
OPEN ENGLISH
$150M
BABBEL
$140M
SPEAK
$90M
CAMBLY
$50M
OTHERS
$500M
* $2B · spent annually on apps that don't solve fluency
Existing apps gamify vocabulary, grammar and reading,
the easy parts. None has built real speaking practice with
personalised feedback, because until recently it was
technically impossible to do on a mobile phone.
85% say English is critical for work
only 25% feel confident speaking it
– Pearson Global English fluency report 2024
*Sorry, but its true. Duolingo is a fun game, but you can spend years on it and not be able to have a conversation in your language of choice.
9:13 PM · Apr 18, 2024 · 1.8M Views
494 733 3K
05 · Why now

This product was impossible to build
three years ago

01
Real-time voice latency dropped below the threshold for natural conversation.
02
Phoneme-level analysis on audio was underdeveloped pre 2022-23 & evolved considerably since.
03
Large language models made conversations realistic and adaptive.
06 · Solution

Stimuler solves English speaking fluency
through Practice, Feedback, and Improvement.

01
Practice
0-latency voice conversations on real-world topics.
02
Feedback
Detailed speech assessment on pronunciation, fluency, vocab, grammar.
03
Improve
A personalised roadmap of targeted exercises.
07 · Traction · Retention

Our paid users keep coming back
across every cohort, and it's accelerating

D1 Retention · All paid · Indonesia
65% ↗17%
48%
65%
Jan '25Apr '26
D7 Retention · All paid · Indonesia
32% ↗15%
17%
32%
Jan '25Apr '26
W4 Retention · Annual paid · Indonesia
38% ↗14%
24%
38%
Feb '25Mar '26
M3 Retention · Annual paid · Indonesia
37% ↗9%
28%
37%
Jan '25Jan '26
4.69
★★★★★
Jan '25·3K reviews
4.80
★★★★★
Aug '25·33K reviews
Now
4.88
★★★★★
Feb '26·95K reviews
08 · Thesis

A generational company in this space
will be built on two pillars

Own Technology

Build in-house, own the voice-to-voice, assessment, and data flywheel layers.

Tech moat
Own Distribution

Acquire users at population scale through organic reach, not paid ads alone.

Distribution moat
09 · Pillar 1 · Own Technology

We built the best speech understanding
system for non-native English speakers

Lower WER than Whisper
42%
WER reduction vs. Whisper · Indonesia benchmark
  • 10× fewer parameters than the largest model matched
  • ~800ms P95 latency in production
Research submitted to Interspeech 2026: "Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR"
Word Error Rate · India, Indonesia, LatAm
% · lower is better
India WER (%) Indonesia WER (%) LatAm WER (%)
Source · Internal benchmark vs. published WER for Whisper, AssemblyAI Universal-3-Pro, NVIDIA Parakeet.
10 · Pillar 1 · Assessment

We assess speech the same way
the world's best exams do, then go further

World's best exams measure
The same four dimensions.
  • 01
    Grammatical Range & Accuracy
    Varied, correct grammar — or repeating the same safe patterns?
  • 02
    Lexical Resource
    Is your vocabulary precise and appropriate for the context?
  • 03
    Fluency & Coherence
    Can you sustain natural, flowing speech without breaking down?
  • 04
    Pronunciation
    Are your sounds, stress, rhythm, and intonation clear to a listener?
Calibrated by actual proficiency examiners.
Stimuler also does
What exams can't.
Measure
where they are right now.
Track
whether they're improving.
Decide
what they should do next.
11 · Pillar 1 · The Agentic System

The first self-improving agentic tutoring system
for spoken English

In every session:
Orchestrating Agent
Unified self-improving tutoring core
01
Conversation
Natural spoken interaction and practice.
02
Feedback
Real-time corrective and formative guidance.
03
Exercise Generation
Personalised drills and speaking tasks.
04
Pedagogical Decisions
Adaptive curriculum and teaching choices.
And this loop has more data to learn from than anyone else in the space.
700K+
Speeches recorded daily
6,000+hrs
User interaction data daily
12 · Pillar 2 · Own Distribution

We built an in-house media network
to acquire users at population scale

Monthly organic views
Millions / month
Jan 2024 → Apr 2026
80M 60M 40M 20M 0
80M+ views / mo
Early '24 Mid '24 End '24 Mid '25 Apr '26
The MATS network today
01
50+
TikTok Channels
Across Indonesia, LatAm, Europe & India
02
~40
Videos / Channel / Month
Consistent publishing cadence across all channels
03
2,000+
Videos / Month
Total network output
04
80M+
Organic Views / Month
Without spending a dollar on ads
The goal Grow this system to population scale, so every person in our target demographic sees a Stimuler video organically every week, without spending a dollar on ads.
13 · Pillar 2 · Economics

This has allowed us healthy economics
while scaling 5.5×

01
~50%
of revenue is organic
No paid acquisition dependency. Users find us, not the other way around.
02
$10–$15
CAC, flat through 5× growth
Paid reach is supplementary, not the engine. Every new dollar of revenue comes with the same acquisition cost.
MRR vs CAC · Mar '25 → Apr '26
MRR ($K) CAC ($)
$400K $300K $200K $100K $0
$30 $22 $15 $7 $0
5.5× MRR · flat CAC in 14 mo
Mar '25 Jun '25 Sep '25 Dec '25 Feb '26 Mar '26 Apr '26
14 · Traction · Indonesia

Indonesia proves that the
playbook works

Indonesia ARR · Jan 2024 → Apr 2026
$1.5M $1M $500K $0
Jan '24 Jul '24 Jan '25 Jul '25 Jan '26 Apr '26
Stimuler #1 in Top Education Apps · Indonesia iOS App Store
15 · Target

Targets for the next 18 months

01

Scale revenue to $25M+ ARR by 2027

Scale further in Indonesia and become #1 in LatAm using the same playbook.

02

Solve longer-term retention & kick off PLG

Get M12 retention to 40% across all our geographies and start product-led growth efforts.

03

Deploy entire voice stack in production

Ship our voice-to-voice architecture in production, validated by published papers.

16 · Backed By

Backed by $3.75M from leading
India & global investors

Lightspeed India
SWC
M Venture Partners
Rebright Partners
Grad Capital
Force Ventures
Operators Studio
&
A group of angels
Top consumer founders
& product operators
And a group of angels

Stimuler

The Highest-Rated Mobile App for English Speaking Practice. Globally.

Akshay AkashCEO & Co-Founder
As of May 15, 2026
A · Appendix · Internal research

Users measurably improve on Stimuler

↑50%
Words per utterance
15.3 → 23.5 d = +0.44 ★★★
↑13%
Words per minute
77.7 → 88.0 d = +0.45 ★★★
+15 pts
Pronunciation accuracy
58.6 → 73.6 d = +0.57 ★★★ (large)
↓57%
Vocab-upgrade gaps
8.9 → 3.8 d = −0.18 ★★★
1,568 users · 1.1M+ scored speech events · 14 of 15 fundamental metrics replicated across 2 independent cohorts · Unscripted speech only
B · Appendix · Academic validation

Independently validated across 2 continents

3 peer-reviewed academic studies by 4 institutions · zero Stimuler funding or involvement

🇪🇨 Ecuador · Conducted by the Ministry of Education Study 1 of 3

+23.1 pts proficiency improvement

120 middle-school students, 3 months. Pre/post-tests across grammar, listening, reading, and speaking. Improvement was significant at p < 0.001 across all 4 skills. No difference between age groups (10–12 vs 16–18) — the app worked for every cohort.

Quasi-experimental p < 0.001 3-month duration 4 skills tested N = 120
Gallo, Romero, Pérez & Marcillo (2026). Impact Research Journal, Vol 4. Universidad Técnica Estatal de Quevedo & Ministerio de Educación del Ecuador. Open PDF ↗
🇮🇩 Indonesia · Largest sample Study 2 of 3

53.8% of speaking variance explained

149 students across 2 vocational schools. Stimuler usage alone had a statistically significant effect on speaking skill (t = 12.26, p < 0.001). Combined with vocabulary mastery, the model explains R² = 0.538 of student speaking-skill variance (F = 84.96).

Multiple regression p < 0.001 N = 149 2 schools
Denistiani, Asipi & Nopiyadi (2025). Tomorrow's Education Journal, Vol 2. Institut Prima Bangsa, Cirebon. Open PDF ↗
🇮🇩 Indonesia · Controlled experiment Study 3 of 3

~4× improvement vs. traditional teaching

38 vocational students split into Stimuler and control groups for 4 weeks. Stimuler users gained 15.1 points vs only 3.9 for the control group. 64% of Stimuler users scored "Excellent" — 0% in the control group did.

Control group N = 38 Pre/post-test 5 skill dimensions
Denistiani (2025). MATCHA Journal, Vol 1. Institut Prima Bangsa, Cirebon. DOI: 10.70152/matcha.v1i1.135. Open PDF ↗