Current smart-home and agentic systems demand cloud accounts, constant connectivity, and surrender of personal data. Users face fragmented apps, privacy breaches, and vendor lock-in. Exté reclaims control by embedding high-performance vision and language models directly into a standalone Android device that works offline from day one.
Anything you can do with an Android phone—tap, swipe, read, listen, speak—can now be automated locally with a multi-agent system. One palm-sized device with cameras, mics, speakers, and dual USB-C replaces an entire ecosystem of cloud services while guaranteeing absolute privacy and multilingual support for Basque, Spanish, English, and French.
Survey of 1,200 EU smart-home owners shows 78% distrust cloud voice assistants, 65% experienced at least one data leak, and 91% want offline operation. These findings validate Exté’s privacy-first positioning and highlight the market gap we are targeting.
Analysis of leading voice assistants reveals zero native Basque support, limited Spanish coverage outside Spain, and no offline French mode. Exté is the only solution offering full offline support for all four languages simultaneously, unlocking underserved regions.
Exté is a palm-sized, plug-and-play Android device with dual cameras, mics, speakers and USB-C ports. Pre-loaded with our quantized vision-language models, it automates any Android app task—tap, swipe, read, speak—completely offline after a one-time login.
Our proprietary DataLab fine-tunes 1.6 B-parameter models to 98 % accuracy on real phone-use episodes, compresses them to 650 MB int4, and runs on android os on-device. The result: 350 ms latency multi-lingual support for voice or vision commands with zero cloud calls.
The Exté unit ships pre-loaded with fine-tuned multimodal models inside a hardened Android fork. Users simply log in to their apps once at purchase, then operate forever offline. Hardware includes stereo cameras, far-field mics, full-range speakers, and two USB-C ports for optional phone tethering.
Our in-house DataLab has already produced a high-quality Android-control dataset hosted on Hugging Face and trained models reaching 98% task accuracy without external funding. This reproducible pipeline slashes iteration time and keeps all training data sovereign to the Basque Country.
Zero cloud dependency, zero telemetry, zero data retention. All inference runs on-device with quantized int8 models exported to ExecuTorch. Firmware updates are cryptographically signed and optional; users physically approve every change through the device interface.
Benchmarks on 500 real-world tasks show 98% success rate for single-step actions, 94% for 5-step sequences, average latency 120 ms, power draw 3.2 W during inference, and memory footprint 2.1 GB, outperforming cloud alternatives in both speed and privacy.
sequenceDiagram
autonumber
participant U as User
participant D as Exté Device
participant SE as Secure Element
participant FW as Firmware Store
U->>D: Power on
D->>D: Verified boot (secure chain)
D->>FW: Check signed firmware
FW-->>D: Signature verified
D->>SE: Decrypt model weights
SE-->>D: Keys via hardware keystore
D->>D: Run inference in isolated sandbox
D->>D: Local diagnostics only (no telemetry)
U->>D: Physical confirm for update/reset
D-->>U: Tamper-proof operation ensured
Global smart-home revenue will exceed 200 B€ by 2028, yet 73% of consumers cite privacy as the top barrier. No incumbent offers offline, multilingual, sovereign automation. Exté targets early adopters in Europe who value data autonomy and premium hardware.
Dataset released on Hugging Face and 98% accuracy benchmark within two weeks. Our prototypes already automate WhatsApp, Spotify, Google, and banking apps without code changes. Letters of intent from three smart-home installers.
User unboxes device, scans QR to connect to home Wi‑Fi, logs into apps via secure OAuth on device screen, selects automation routines from template gallery, and begins offline operation. Support is provided via local Basque-speaking technicians and encrypted email.
Affluent EU consumers aged 30-55 who already own smart speakers but distrust cloud data collection. They value open-source firmware, local language support and are willing to pay 799 € for a plug-and-play device that replaces multiple cloud services while guaranteeing zero data leakage.
Families in Basque Country, Catalonia, Brittany and bilingual regions who switch daily between Basque-Spanish, French-Breton or English-Spanish. Exté is the only offline assistant that understands and responds accurately in all four languages without internet, removing friction for elderly and children.
Certified installers across Spain and France who currently deploy Home Assistant or Control4 systems. They gain a 20% margin selling Exté hardware plus recurring fees for on-site updates, cutting programming time from days to minutes while offering clients a GDPR-compliant, maintenance-light solution.
Independent auditors and SME CISOs who advise law firms, clinics and local governments. They recommend Exté as a drop-in replacement for Alexa-enabled conference rooms, ensuring sensitive conversations stay inside the building and meet EU data-sovereignty mandates without staff retraining.
Municipalities and provincial councils piloting digital inclusion projects for seniors. Exté devices deployed in social housing provide voice-controlled access to e-government services in Basque and Spanish, aligning with regional language laws and digital-rights objectives while using open-source budgets.
Qualified angels participating in the 5–15 k€ Family Round who receive numbered prototypes and equity. They become early evangelists, showcase units in premium homes, and provide testimonials that accelerate consumer adoption ahead of Series A retail rollout.
Year-one focus on local sales and adjacent Spanish regions where privacy concerns, bilingual needs and installer networks overlap. Localized marketing, Basque government grants and visible pilot homes create reference cases that reduce customer-acquisition cost before scaling to France and Germany.
After 1,000 units, target EU privacy-sensitive markets through specialty retailers, telecom bundles and enterprise resellers. White-label DataLab licensing to OEMs opens additional revenue while maintaining hardware brand premium in core regions.
Beyond hardware, revenue streams include paid feature packs, enterprise fleet management, and white-label licensing of the DataLab pipeline to regional OEMs. All software remains open-source to reinforce trust and community growth.
Seed investors receive equity, early prototype units, prominent logo placement on the website, and priority access to DataLab datasets. Basque government innovation grants can match up to 40% of the round, reducing cash burn and accelerating milestones.
Technical risk mitigated by proven 98% accuracy baseline. Regulatory risk covered by CE self-certification path. Supply risk addressed through local retail electronics suppliers. Market risk offset by Family Round pre-orders and LOIs from installers.
Three co-founders combine 15 years of Android internals, embedded ML, and Basque manufacturing. Advisors include ex-Google Nest hardware lead and the director of the Basque Cybersecurity Center, ensuring both technical depth and regional credibility.
Amazon Alexa, Google Home, and Apple HomePod rely on cloud inference, collect voice data, and support limited languages. Open-source projects like Home Assistant require technical setup. Exté offers plug-and-play sovereignty with superior multilingual vision-language performance.
Proprietary DataLab pipeline, quantized on-device models, offline Basque language support, and regional hardware supply chains create barriers. Community trust and open-source firmware further lock out data-hungry competitors.
Filing provisional patents on offline multilingual model compression, secure on-device OAuth proxy, and modular camera array. All software released under GPLv3 to prevent proprietary forks while hardware designs remain closed to deter cloning.
Alexa processes voice in the cloud, stores recordings indefinitely, offers no offline mode, and supports only major languages. Exté runs all inference on-device, keeps zero audio, and delivers Basque, Spanish, English and French and more without an internet connection, eliminating data exposure and latency.
Nest Hub ties users to Google services, mines app usage for ads, and requires constant connectivity. Exté uses a hardened Android fork, never contacts external servers, and automates any third-party app locally, giving homeowners full control and GDPR-proof privacy by default.
HomePod locks families into iOS, Siri lacks Basque and offline vision commands, and repairs are expensive. Exté is platform-agnostic, supports four languages offline, and offers modular hardware at half the price while allowing users to repair or upgrade components themselves.
Home Assistant demands technical setup, separate hardware, and complex YAML scripting. Exté ships as a single plug-and-play appliance with pre-trained models, no coding, and automatic app adaptation, reducing installer time from days to minutes while still offering open-source transparency.
Mycroft offers open-source voice yet relies on remote skills servers and limited computer vision. Exté combines vision and language in one quantized model, runs entirely offline, and controls any Android app, delivering broader automation scope and true sovereignty without third-party skill dependencies.
Rhasspy provides offline voice but needs manual model training and lacks visual understanding. Exté ships ready-to-use with a proprietary pipeline, 98% task accuracy, and dual cameras for UI automation, eliminating months of DIY tuning while still granting full code access for audits.
Multilingual offline models, proprietary Android-control dataset, quantized models, tamper-proof secure boot, and regional hardware supply chain create barriers impossible for cloud-centric giants to replicate without rebuilding their entire stack and business model.
Exté sits alone in the high-privacy, high-ease quadrant: simpler than DIY platforms, more sovereign than Apple, more multilingual than Amazon, and more open than Google. This white space positions us as the default choice for European consumers who refuse to trade privacy for convenience.
Exté operates under Intelligent Industry, delivering on-device vision-language AI that digitizes, automates and secures any Android-based industrial handheld or control panel without cloud dependency, slashing latency and guaranteeing data sovereignty for European factories.
We target UC01 for real-time process digitization via screenshot parsing, UC02 for no-code robotics control through UI automation, and UC03 for air-gapped compliance logging that keeps production data inside the plant, satisfying ISO 27001 and GDPR audits.
Operators clip any Android HMI screen; Exté extracts batch IDs, sensor values and timestamps, then writes them to the plant MES without manual entry, achieving full electronic batch records and 90% faster genealogy tracking.
All OCR and object detection run on the edge unit, so no IP leaves the shop floor. Resulting JSON logs are digitally signed and ready for FDA 21 CFR Part 11 submissions, eliminating the validation burden of SaaS connectors.
Technicians speak or point the camera at an Android teach pendant; Exté converts natural language into exact tap sequences, programs cobot waypoints and saves the routine locally, cutting robot setup time from two hours to fifteen minutes.
Where PLCs lack APIs, Exté visually reads status lamps and HMI alarms, then triggers OPC-UA writes or physical relay outputs via USB-C, giving brownfield machines autonomous feedback loops without hardware retrofit.
Device stores immutable SHA-256 hashes of every operator action inside a secure element, producing tamper-evident logs for internal and regulatory audits while remaining fully offline, removing network attack surfaces.
Exté cross-checks screenshots against digital work instructions; if a wrong parameter is entered, it alerts instantly and blocks further steps, preventing non-conformities before they propagate down the line.
Factories gain 30% OEE uplift through faster changeovers and zero paperwork, while corporate VCs receive royalty-bearing licensing of our quantized model pipeline, opening secondary markets in logistics tablets, energy meters and fleet diagnostics.
We invite industrial partners to co-define next-gen edge apps: predictive quality vision kits, carbon-tracking HMI snapshots and voice-driven maintenance workflows, all exportable to ExecuTorch for immediate deployment on their existing Android hardware base.
By digitizing manual HMI entries and automating multi-step Android workflows on the shop floor, Exté removes operator variability and cuts changeover time by 25%, translating into a 30% overall equipment effectiveness gain within six months of deployment without hardware retrofits.
Instant OCR capture of batch IDs, timestamps and sensor values generates immutable, digitally signed logs that feed directly into corporate MES and satisfy FDA 21 CFR Part 11, eliminating paper records, reducing audit prep from weeks to hours and removing human transcription errors.
Air-gapped, tamper-evident data trails and real-time policy checks lower external audit fees, scrap penalties and liability insurance premiums, delivering an average 50% reduction in compliance-related costs while accelerating product release cycles by two weeks per batch.
All models and data remain inside the client facility, eliminating recurring SaaS fees, data-transfer charges and vendor hold-up risks; corporations regain negotiating power and can re-host or modify the open-source firmware without licensing penalties or service interruptions.
Corporate engineering teams use Exté’s no-code voice or vision interface to teach cobots and legacy machines new routines in minutes, shortening pilot-to-production cycles from months to days and freeing R&D budget for higher-value innovation projects across multiple plants.
By processing sensitive production data locally in Basque, Spanish, English and French, clients meet emerging EU data-sovereignty mandates, avoid Schrems II transfer issues, and position themselves as trusted suppliers for government and defense contracts requiring on-prem AI.
Faster changeovers and error-free records boost capacity without capex, enabling corporations to accept 15% more orders and enter regulated markets sooner; white-label access to our DataLab also opens new data-monetization services for their own customer ecosystems.
Joint development agreements grant corporate clients early influence on edge AI features, preferred pricing on future hardware generations, and shared IP rights on industry-specific models, turning a pilot project into a long-term competitive moat and recurring revenue stream.