MOSTRAmi is an AI museum guide born on Android, built in Kotlin and Compose, powered by on-device language models and a secure proxy client. It turns a photo of an artwork into an audio explanation in the visitor's language, preserving the museum's editorial authority as its anchor of truth, and visitor choice as the sole source of consent. It is a consumer product serving as the wedge for a broader B2B cultural platform. We start in Europe because it is here that privacy, accessibility, and algorithmic transparency have already become competitive advantages rather than legal overhead.
There is a subtle but persistent gap between what a work of art wants to say and what a visitor, standing before it, can actually grasp. The reasons are familiar to anyone who has visited an exhibition: labels are brief due to physical space limits, audio guides are generic to suit everyone, curators are rarely available, and the smartphone in the visitor's hand does many things but not yet this. Our thesis is that a personal, contextual, multilingual conversation grounded in curatorial content is no longer a luxury but a baseline expectation. The first company to deliver this in a credible, sustainable, and privacy-respecting way will establish a highly defensible position.
MOSTRAmi is built on this thesis. It does not replace the museum: it amplifies it. It does not replace curators: it translates, contextualizes, and guides, leaving decisions about authoritative content in their hands. The product is both a technological step and a positioning choice. This briefing outlines both dimensions without mixing marketing with engineering.
"The right product at the right time: smartphones are now powerful enough to run local LLMs, and Europe is ready to reward applications that treat visitor data as a fiduciary trust."
Most cultural institutions combine static labels, single-exhibit custom apps, and rented hardware audioguides that run parallel to the visitor's smartphone. The result is a fragmented experience: three touchpoints for three different target groups, no shared memory, no relationship with the visitor after they leave the building, and no structured feedback for the curator. Conversely, generalist AI platforms offer conversational answers but lack connection to the museum's official archives, making them a source of institutional distrust.
MOSTRAmi is designed as a neutral layer: the visitor opens it at the entrance, uses it in front of the art, and reviews it afterward. It does not require replacing existing systems, but it provides a clean point of integration: official PDF maps, curatorial texts, and dashboard logs. This neutrality serves as a B2C wedge that can transition into a high-margin B2B platform over the next two to three years.
A common mistake for cultural startups is starting with direct museum sales. Sales cycles are long, decisions are consensus-driven, and products age while waiting for pilot programs. We chose the opposite path: build a consumer app that works for tourists, students, and families, build usage metrics, and only then approach museum administrations with real data rather than slides.
Begin structured closed access only after signing, cloud canary and exact-device gates are green, then expand the cohort gradually. Goal: validate repeat use, voice and camera reliability, device compatibility, support load and cloud cost before any open beta.
Complete one scoped exhibition deployment, measure visitor outcomes and test a commercial offer. Use the evidence to decide whether Android is ready for an open beta or limited public Play release.
Scale Android, museum tooling and accessibility only after retention and reliability are proven. Evaluate iOS, WeChat and AppGallery from measured demand rather than fixed launch dates.
The app is written in Kotlin with Jetpack Compose for the UI layer, runs on Android 8+ (API 26), and aligns with the targetSdk 36 requirement for 2026 Google Play releases. The modular structure separates camera, AI, audio, navigation, billing, AR, and export systems. Every sensitive feature (camera, microphone, location, cloud API) passes through a single point of control. On-device Gemma keeps visits private and resilient; Gemini Cloud and Gemini Live are used when they add clear quality, depth, or voice-interaction value.
Single-activity, declarative navigation, Material 3 styling, accessibility-first design. Uses Compose BOM, KSP2 code generation, and Coil image loading.
Gemma model served through LiteRT-LM (Google AI Edge runtime). Built-in package download and status manager accessible in both Standard and Expert profiles. Local RAM management with structured cloud routing.
On-device label text extraction, with optional ARSceneView Compose integrations for context overlays. Camera-level face obfuscation before any data transfer.
The cloud text path uses Gemini 3.6 through a structured generator, deterministic citation validation and a separate semantic/safety verifier. Gemini Live remains a dedicated native-audio path. Production deployment, quota controls and physical-device behavior remain release gates; telemetry is consent-gated.
History, artwork records, user preferences, and offline maps are stored locally. Artwork photos are not saved without opt-in. Complete history export and deletion available in settings.
Freemium plan with scan quotas, premium voices, and B2B packages. Implements rate limits, purchase restoration, and subscription structures compliant with 2026 Play Store guidelines.
After the Android closed beta and first exhibition evidence, two optional expansion channels can lower onboarding friction for casual visitors. Messaging can turn existing chat clients into zero-barrier cultural guides; TourSync can support guides and groups. Neither channel has a fixed release date before the core beta and pilot gates close.
Casual or international tourists often resist installing custom apps. Our roadmap response is the **Clio Messaging Proxy**:
Tour guides rent expensive physical radio sets for their groups. Our **Clio TourSync** feature will digitalize and optimize this interaction:
Language models, computer vision, and text-to-speech are commodities, available to anyone paying for API calls. Defensibility lies in creating a product that earns and maintains institutional trust in the European market, where trust has a clear barrier to entry. Our competitive advantage relies on four layers: each is replicable individually, but they become highly defensible when combined.
Diagnostics are deactivated until consent is given, photos are not saved without opt-in, and cloud AI queries are proxied. This is more than compliance: it is a pledge that European visitors value.
On-device Gemma is the default. This keeps unit costs low, supports visits in historical basements without signal, and is the only setup some institutions will permit.
Privacy policies, Play Data Safety forms, AI transparency documentation, and in-app account deletion options are already in place, making the app ready for EU audits under the AI Act.
The project bridges the gap between museum administration requirements and modern Android engineering. This dual understanding is a key early-stage asset for our team.
This is the current product state as a technical reviewer would see it: an unsigned 0.8.0 source candidate with passing tests and build, plus the current signed v0.7.0 private-beta APK, which is the only distributable build for authorised testers and is also retained as the rollback artifact for future v0.8 qualification. The checkpoint is dated August 14 and does not claim public deployment.
To address the entire European tourist demographic and enable institutional licensing contracts, the MOSTRAmi ecosystem plans native porting to all three major store platforms:
Invite-only Android testing begins after the release-readiness gate. Public Play distribution remains gated by beta evidence, backend reliability, billing QA and the wider device matrix.
Native porting in Swift and SwiftUI. Aligned with Apple CoreML neural engines and LiDAR tracking in AR.
Native ArkTS/HarmonyOS NEXT porting to engage with Chinese and Asian tourist flows visiting Europe.
Native Mini Program for Chinese tourists in Europe. Museum QR code opens an instant Mandarin guide with zero downloads. Full bypass of GMS blocks on Huawei and Xiaomi devices.
Fully validated Business Model Canvas, Lean Plan, and 10-slide Pitch Deck with metrics, monetization strategy, and growth projections.
OPEN YC EVIDENCE & RELEASE STATUS → ⚡ OPEN SLIDE PITCH DECK →This page is not a request for a term sheet. It is an opening of dialogue with investors, developers, and museum operators who want to learn about MOSTRAmi's path. If you would like to ask questions or give feedback, let's talk.
Email Fabrizio Degni →