AI that sees, listens and rings up the sale.

GateSense turns Ubiquiti UniFi cameras into a system that recognises faces and number plates, answers spoken questions about what it sees, and opens a door only when two independent signals agree.

• Face and number plate recognition plus custom detections such as the postman, a parcel on the doorstep or a pet in the yard.
• Ask the camera anything. "Did the postman come?" or "Is the garage door open?" in spoken Slovenian, answered by a local vision model.
• Two-factor by design. Face plus voice, or plate plus phone at home, before anything unlocks. Every opening sends a notification.
• Nothing leaves your network. No images or audio go to the cloud.
• With Home Assistant or without it. A standalone web admin with its own rule engine is included.

Built with UniFi Protect, Frigate, Whisper, Piper, MQTT, Home Assistant and Docker on a GPU server.

DeskSense is a voice agent that answers questions, checks availability and books appointments in English and Slovenian. Callers can interrupt it mid-sentence, the way they would a person.

• Natural interruption. New speech stops the current reply and the agent picks up the new thread.
• Bookings that hold. Nothing is booked on a vague "yes": the agent reads the proposal back and waits for explicit confirmation, and slot locking prevents double bookings.
• Local speech. Recognition and synthesis run on your own machine, with optional premium voices per language.
• Your model, your knowledge. Connects to the language model you already run and answers from your services, catalogue and calendar.
• Staff calendar included. Month, week and agenda views of confirmed appointments.

Browser voice calls work today. Phone lines are the next adapter on the roadmap.

Built with Pipecat, WebRTC, FastAPI, PostgreSQL, Redis, Silero VAD, faster-whisper and Piper.

SellSense is a mobile-first marketplace where the photo is the listing form. A local vision model recognises the item and writes the title, description, category and a suggested price. The seller adds a location and publishes.

• From photo to listing in seconds, with automatic categorisation and a realistic price suggestion.
• 100% local AI. No cloud vision API, no per-listing cost and no photos shared with a third party.
• Instant, typo-tolerant search with filters, even across thousands of listings.
• Built to scale. Photo galleries are optimised and served straight from object storage.

Built with Next.js, FastAPI, PostgreSQL, Meilisearch, S3-compatible storage and Gemma 4 on VLLM.

TillSense is a restaurant point of sale that runs in the browser or as a native desktop app from the same code. Menus, tables, orders, payments and receipts live in one place.

• Tables, orders and colour-coded menus designed for fast tapping during service.
• Totals calculated on the server, so tax and sums are identical on every terminal.
• Cash and card payments with a PDF receipt at the end of each transaction.
• Installs like a desktop app or opens as a web page, whichever suits the venue.

Built with React, Redux Toolkit, TanStack Query, Tailwind CSS and Tauri 2.

TillSense Go is an Android cash register for hospitality, shops and service businesses. It keeps issuing receipts without an internet connection and comes with FURS integration.

• Works offline on an Android phone or tablet.
• FURS integration for Slovenian fiscal verification of receipts.
• One app for bars, shops and services.

SeatSense is a ticketing platform with live seat maps, season tickets, QR tickets and a gate scanner that runs on any phone. Each organiser gets a separate instance.

• No double-selling. Seats are held atomically for five minutes while a buyer checks out, with a database guarantee behind it.
• Live seat map. Every visitor sees holds, releases and sales as they happen.
• Venue designer with AI floor-plan scan. Upload an image of the hall and a vision model drafts the sections for you to review.
• Signed QR tickets for single events, whole seasons and free entries, checked by a camera scanner at the gate.
• One venue, many events. Reuse a seat map with different ticket types and price categories per event.

Built with ASP.NET Core 8, React, TypeScript, PostgreSQL, Redis and server-sent events.

SignSense sends a one-time link, lets the signer read the PDF and draw a signature on any device, then returns the signed document by email and webhook. It runs on your own servers.

• One-time 256-bit links that expire, are stored only as hashes and die the moment the document is signed.
• Vector signature in the PDF, with an optional PKCS#7 digital signature on top.
• Append-only audit trail of every upload, view, signature and rejection.
• Plugs into your systems. A simple API and HMAC-signed webhooks tell your software when a document is done.
• Documents stay with you. Private storage, rate limiting and strict security headers out of the box.

Built with ASP.NET Core 8, React, PDF.js, PostgreSQL and S3-compatible storage in Docker.

Sirisense Compute is our own multi-GPU workstation for deep learning, image and video processing, simulations and other heavy workloads.

• 4 × NVIDIA RTX 3090 with 24 GB of memory each.
• 64-core AMD Threadripper Pro and 256 GB of ECC memory.
• 6 TB of NVMe storage for large datasets.

FieldSense combines compact sensors with agro-climatic forecasting models, so growers can act on the risk of disease and pests in time.

• Air temperature, humidity, light and soil measured right next to the plants.
• For fields, vineyards and orchards, with RFID tags that track produce through the cold chain.
• Timely warnings in a web app, for more targeted plant protection.