Built and running.
Selected projects from Axiforge Systems — spanning cognitive software, proprietary mechanical R&D, and autonomous systems.
LED Strip Production Line
Embedded Systems · AI Quality Control · Industrial Automation
Fully operational LED strip production line built on custom embedded control systems and AI-powered quality inspection. Demonstrates Axiforge Systems' capability to deliver end-to-end industrial automation — from embedded firmware to AI inference on the production floor.
Memvo — AI Language Immersion Platform
Memory Palace AI · LiveLens Camera · LensCompanion Device
The world's first platform combining Memory Palace AI, real-time camera immersion, and offline hardware learning into a single English acquisition ecosystem. Three modules — web, mobile PWA, dedicated device.
Visit memvo.pl →Novel Transmission Mechanism
UK patent application pending · IPO 2025
A novel mechanical transmission concept developed by Axiforge founder Matthew Foster. The mechanism achieves improved torque density and efficiency in a reduced form factor — with direct applications in UAV propulsion, actuator systems, and mobile robotics. Prototype validated. Full technical details available under NDA.
Licensing enquiries →Memvo LensCompanion
LensCompanion is a dedicated English learning device that runs entirely offline — no cloud, no subscription for AI, no privacy concerns. Built on Raspberry Pi 5 with Sony's IMX500 image sensor, it is the only language learning device in existence that uses on-chip neural network acceleration for real-time object detection.
Raw video frames are processed directly on the camera sensor's built-in AI chip and never transmitted — not to the cloud, not even to the host processor. This is privacy-by-hardware architecture, not a policy promise.
| Procesor | Raspberry Pi 5 (16GB RAM) |
| Kamera | Sony IMX500 — on-chip AI accelerator |
| Detekcja | EfficientDet-Lite0 · 30fps · 80+ klas · 0% CPU |
| STT | Whisper.cpp small.en · offline · ~250ms latencja |
| TTS | Piper TTS en-GB · offline · ~150ms latencja |
| Dialog LLM | Qwen2.5 3B Q4 · offline · 5–8 tok/s |
| Vision LLM | Moondream2 Q4 · offline · opis sceny w ~4s |
| Połączność | WiFi sync z memvo.pl · BT słuchawki |
| Autonomia | ~4–5h (powerbank 10Ah) lub zasilanie ciągłe |
| Koszt API | $0.00 per sesja (wszystko lokalnie) |
Three Modes of Operation
HOME MODE
Always On, Always Learning
Passive 24/7 operation. The IMX500 detects objects in your environment — a laptop, a coffee cup, a book — and the device quietly whispers the English name with IPA pronunciation. Every 30 seconds, PoseNet detects your body posture and teaches action verbs: "You are sitting", "You are reaching", "You are walking." Power consumption: ~3W. Silent operation.
STUDY MODE
Your Pocket Language Tutor
Active vocabulary sessions using words from your Memvo Memory Palace. Three GPIO buttons: Know / Don't Know / Next. Qwen AI explains words you struggle with. Results sync to memvo.pl after each session.
WALK MODE
AI Narrates Your World
Take the device on a walk. Moondream2 describes what the camera sees every 8 seconds, weaving in your palace vocabulary naturally. Push-to-talk button for two-way conversation with Qwen AI. Speech recognition via Whisper.cpp — all processed locally.
☀️ MORNING REPORT
Every morning at 7:00, LensCompanion delivers a personalised English summary: which objects you saw most, new words encountered for the first time, active hours, and an AI-generated language tip from Qwen for your most-seen word.
Custom Oxford 3000 Detector — In Development
A custom IMX500 model trained on Oxford 3000 vocabulary will replace the generic COCO detector, enabling recognition of 400+ Oxford words directly on the camera chip.
Humane AI Pin: $700. Failed.
Rabbit R1: $200. Failed.
Both failed because they tried to do everything for everyone.
LensCompanion does one thing: English acquisition.
It does it better than any device ever built.
Part of the Memvo ecosystem — memvo.pl
Enquire about LensCompanion →Live demo — Memvo
Try the web platform
Fully interactive — embedded directly from production.