LMC Tech
Charity web platform
TimeOut Recovery
Three regional services, one platform: referral routing, resource library, regional editors and a signed data processing agreement, for a lived-experience gambling recovery service.
Our work / Published Android app
LODORi
Tell it what you need and LODORi builds a guided session for that moment: script, voice, ambience, mixed to broadcast loudness. Built and shipped by us from first commit to a public Google Play listing with paying subscribers.
~25s
to generate a five-minute personalised session
6
countries at 100% rollout on Google Play
180
commits from empty repo to public release
4
TTS providers in a fallback chain, so a vendor outage never blocks a session
The brief
Meditation for the moment you are actually having.
Most meditation apps are libraries. LODORi is a generator. You say what you need, how long you have and how you feel, and it writes a script, voices it, lays it over ambience and hands you a session that did not exist a minute ago.
That only works if it is safe. Someone in crisis must never receive a chirpy breathing exercise, so the safety layer is not optional and cannot be skipped.
- React Native (Expo bare) Android app with Apple and Google sign-in, journaling, ratings, favourites, sharing and progress tracking.
- FastAPI backend with Celery workers for script, audio and background generation; PostgreSQL, Redis, Cloudflare R2 with signed URLs.
- Safety-first pipeline: self-harm, crisis and medical pre-check before generation, a post-generation validator and a prompt-injection guard, with a documented policy and test corpus.
- Provider-agnostic AI: LLM chain Groq, Gemini, then self-hosted Ollama; voice chain Google Neural2, Groq Orpheus, ElevenLabs, OpenAI. Unconfigured providers are skipped automatically.
- Freemium business logic: daily limits, rewarded-ad unlocks, 14-day trial, monthly and annual subscriptions via RevenueCat on mobile and Stripe on the web to avoid the store cut.
- Next.js marketing site, PostHog analytics, Sentry monitoring, AdMob approved, and the Play Store listing, screenshots and ad creatives.
Under the bonnet
A product, not a prototype.
The generation path is script, then text-to-speech, then ffmpeg mixing at -16 LUFS voice over a -34 LUFS bed, so every session sounds like it came from the same studio. A voice cache keyed on script, voice and provider keeps TTS spend down without leaking one voice into another.
Everything runs on one VPS behind Cloudflare with systemd and nginx. Store billing is handled by RevenueCat so the app never sees a card number.
- React Native
- Expo
- FastAPI
- Celery
- PostgreSQL
- Redis
- RevenueCat
- Stripe
- AdMob
- Next.js
- ffmpeg
What it shows
- We build and publish native mobile apps, including the store listing, billing and compliance.
- Generative AI shipped to the public with a real safety architecture, not a disclaimer.
- Resilient integrations: four voice vendors and three language-model vendors, all interchangeable.
- iOS build not yet released; the backend and design are ready for it.
AI social media manager
Social Jen
Give it one idea, a product photo or a video and it drafts, schedules and publishes to nine social networks, writes graded Etsy listings and renders your 3D models. Nothing goes out until you approve it.
E-learning platform
LMC Training platform
Safeguarding, Prevent, drug and alcohol service induction, trauma-informed practice and more, with server-marked knowledge checks, verifiable certificates, employer dashboards and Stripe seat packs.
Next step