second open
← All solutionsSolution · Real-time AI on mobile

AI that answers before the user moves on

Instant feedback on touch, pencil, voice or camera input, streaming that survives a train tunnel, and AI features that get through App Store and Play review. On a phone the model is the easy part. The network, the device and the store are where it breaks.

Who it is for

Mobile products where AI has to react in the moment: learning, fitness, creativity, health and games.

Book a 30-minute callSee where we have done it

What we can build for you

Six things we can put in your product, each backed by work we have already shipped.

01
Feedback on touch and pencil
Handwriting, drawing and tracing checked stroke by stroke, as they happen.
02
Streaming that survives bad networks
Resumable, cancellable and cheap to retry, with nothing lost when the signal drops.
03
On-device or server, decided on numbers
Latency and unit cost decide where each model runs, not fashion.
04
Store review for AI content
App Store and Play submissions for AI features, which fail for undocumented reasons.
05
Nudges at the right time
Reminders scheduled in each user's timezone, one a day, never two.
06
One team, three platforms
Flutter or native, iOS, Android and web, with the backend behind them.

The four tests, for this

Where this kind of feature usually breaks, and what passing each test looks like.

01 · Design
Does it feel right?

Feedback arrives while the gesture is still happening, and a mistake gets a nudge, not a red cross.

02 · Mobile and backend
Most at risk
Does it hold up?

Works on old phones, shared tablets and flaky connections, and fails in ways that look deliberate.

03 · AI, retrieval and evals
Is it right?

The model's judgement is checked against a reference built with experts in your field.

04 · Metering and pricing
Does it pay?

Inference runs on the device where it can, so the cost of a session stays predictable.

Closest track: Build it from zero

How the two weeks go

€6,000, fixed. Written findings on day five; stop there and pay half. You keep everything either way.

  1. Days 1 to 4Profile the feature on real devices and real networks. Measure latency, failure modes and cost per session.
  2. Day 5Findings: on-device against server, the latency budget, and what the stores will object to.
  3. Days 6 to 9The feature built into your app to that budget, streaming and resilient.
  4. Day 10Handover with device test results, the cost model and a release checklist for store review.

Where we have done it

Built by our team · 2023
Educational app for kids
Real-time feedback on touch and pencil input as it is drawn, on iPhone, iPad and Android.
Read the case study
Built by our team · 2025 to now
Voice AI writing app
Live speech on three platforms, with each answer saved before the stream finishes.
Read the case study

Tell us what you are building.

Thirty minutes, no deck. You get a straight answer about whether we are the right people, and we say no when we are not.

Book a 30-minute callpoppaultudor@gmail.com