Inspiration
Meet the future of language learning (and e-learning!). I built one of the most powerful language apps on the planet – entirely without writing a single line of code (100% Bolt magic!).
A problem many people have is being able to learn skills and languages in a quick, fun, interactive and memorable way that they enjoy.
LingoHive solves that. I've built interactive e-learning systems before for industry and have seen the impact they have on the world when done right.
A key inspiration I had was to get my 11 year son involved in Bolt, AI and tech, he helped with this project and also submitted 3 of his own
What it does
- Talk to 42 AI Tutors! Chat naturally (type OR speak!) with 42 unique AI personalities on diverse topics. They speak back (thanks Deepseek & ElevenLabs)! Get real-time grammar & spelling correction in any language.
- Interactive Video Tutors! Engage with AI-powered video teachers (powered by Tavus). It's like having a personal coach anytime.
- Smart, Adaptive Learning: Dynamic difficulty adjustment, personalized quizzes, and engaging lessons that go way beyond the basics.
- Seamless Experience: Full user logins & progress tracking (Supabase backend), smart notifications, recent activity feeds, and a slick permissions system (Free, Registered, Paid tiers – try the demo logins on the login forms).
- Massive Scalability: 12 languages live (I can easily add more!), slick onboarding, and a platform I can instantly adapt to create any e-learning tool imaginable.
- Built for Real Impact (and Value): This app can dent the language learning and so many other markets whilst teaching people skills in ways that really help them learn with ease and fun.
- Professionally designed UI/UX (which I did with my own sketched and Google's new UI tool, Google Stitch). Fully mobile-responsive,
Shoutout to my 11-year-old son who helped me a bit on this app, he's also submitted 3 of his own apps to this hackathon.
I've been working in Product and Tech for 15 years and am an early adopter of many AI tools.
How I built it
- Bolt (not one piece of custom code or any other software dev tool), *Deepseek to generate chatbot message, dynamic personas, hints, spelling & grammar checks and response *ElevenLabs for text to speech and speech to text, *Tavus for Video interactive bots and *Supabase for the authentication, databases to save users progress and edge functions.
Challenges we ran into
- The project started getting too big for Bolt so I had to spend a lot of time optimising the project and dynamically loading content into it to bypass this
- Had to get creative with LLMs to dynamically generate content to reduce hardcoding some elements
- Tavus API was tricky to integrate and also very unique as there are some key thing I needed to do to force the video bot session to end on Tavus's servers
- ElevenLabs voices were getting too excited after a few conversations so I had to create some custom workaround to naturalise them
- I shared many of my tips with the Discord community and it helped quite a few people
Accomplishments that we're proud of
- One of the most powerful, engaging, simple and advanced language learning and e-learning tools in existence
- An app that is so cool that you can have 2 way voice and text conversations with 40+ chatbots
- Dynamic real-time learning content that constantly adapts to each user
- Managed to make this fully with Bolt, not a single line of custom code or other code-gen tools
- This app should have taken 5 developers, 4-5 months and $50k+ to build. I know this and have got it checked out by multiple people to confirm.
What we learned
- How to work around the 'project being too large' and solving that
- How to push Text to speech, speech to text, LLM personas and video bots to their limits
- How to rapidly create interactive lessons for multiple languages (modules, lessons, quizzes,etc), chatbots and video bots
- How to create learning content that adapts to each individual user in real-time
- How to perfect notifications that are slick, engaging and proactive for each different user
What's next for LingoHive
- Improve latency on one of the APIs. I noticed the Deepseek API (used to power the chatbots conversations and personas) slowed down when you use the Netlify deployed URL after I moved the Deepseek API to Supabase edge functions
- Integrate Stripe, I've already done that on my other project submission
- Create lesson content for many more languages and modules and get them to dynamically load in from Supabase for users
Built With
- bolt
- deepseek
- elevenlabs
- google-stitch
- netlify
- react
- supabase
- tavus
- typescript
- vite




Log in or sign up for Devpost to join the conversation.