Projects

What we have built and shipped.

Reference projects from YAPZEK® engineering — each one running in production or in customer pilots today.

YAPZEK® Stream — Closed-network multilingual broadcast for universities
Customer pilot2025 — 2026

YAPZEK® Stream

Closed-network multilingual broadcast for universities

A lecture is spoken in Turkish; students listen in English, German or Arabic on their own phones — with live subtitles and a downloadable transcript — without a single packet leaving the campus.

Deployed in university pilots in Istanbul, YAPZEK® Stream runs the full STT → translation → TTS → subtitle pipeline on a single GPU server inside the institution's network. Listeners join by scanning a QR code; the presenter manages languages, recording and listeners from one panel.

The same deployment pattern is offered to hospitals and medical congresses, public institutions and secure-network environments.

LanguagesTR → EN · DE · AR and more
LatencySub-second end to end
AccessQR / web, no app
NetworkFully closed, offline capable
YAPZEK® Video — Offline video generation platform — end-to-end, long-shot consistent
In production2026

YAPZEK® Video

Offline video generation platform — end-to-end, long-shot consistent

A desktop studio that turns a script into a finished multi-scene video on your own GPU. Storyboard, first-frame images, scene videos, voice and soundtrack are generated end to end, with the same presenter kept consistent across every shot — no cloud, no token cost.

YAPZEK® Video breaks a script into timed scenes, writes a video prompt, an audio prompt and a first-frame image prompt for each, then generates the whole sequence in order. Scenes can start from the last frame of the previous shot or extend the previous video, so a ten-scene, 100-second piece keeps the same face, wardrobe, lighting and camera from the first second to the last.

Everything runs locally: LTX video models with LoRA support, image generation and editing, an upscaler and a timeline editor that places every finished clip in order. The presenter speaks the script with YAPZEK® TTS, so product storyboards, announcements and avatar agents are produced in Turkish and English from the same text.

PipelineScript → storyboard → images → video → voice
ConsistencyLast-frame continuity, scene extension
ModelsLTX video + LoRAs, local image gen, upscaler
Runs onYour GPU, fully offline
YAPZEK® Phone
YAPZEK® Phone
Hardware + AI2026

YAPZEK® Phone

Raspberry Pi and ESP32 AI call bridge

A voice agent on a real phone number — and a hardware bridge that connects any mobile phone to the YAPZEK® voice stack over Bluetooth HFP. Raspberry Pi for the full bridge, ESP32-WROOM-32 for the lightweight edge variant.

YAPZEK® Phone answers calls with the YAPZEK® voice pipeline: ASR, LLM reasoning and TTS, optimized for 8 kHz μ-law telephony. The public demo line runs a factory appointment scenario and is integrated with both a website and a phone switchboard.

The hardware bridge pairs with a phone as a Bluetooth hands-free device, streams call audio to the agent and plays the response back — turning any existing SIM and handset into an AI-answered line with no telco integration.

HardwareRaspberry Pi · ESP32-WROOM-32
LinkBluetooth HFP, Wi-Fi 2.4 GHz
Audio8 kHz μ-law telephony
Demo line+90 850 203 64 79
NxtEdge — The AI Operating System
In production2026

NxtEdge

The AI Operating System

An entire company modelled as a graph — departments, people, agents, systems — with governed agents executing real work under HITL, HOTL or autonomous modes.

NxtEdge is YAPZEK®'s largest platform: a relational organisation core, an AI assistant that queries real databases and writes reports, a visual flow engine with AI-generated nodes, a scheduler, an event bus, MCP-governed tools and a host-agent for local browser, database and inference access.

It ships with frontier and local model support, audit trails, RBAC and PII protection, and is deployed privately with dedicated onboarding.

CoreRelational org graph
GovernanceHITL · HOTL · HOOTL
ToolsMCP, unlimited integrations
ModelsFrontier + local
talk.yapzek.ai — Self-service voice agents for any website
In production2025

talk.yapzek.ai

Self-service voice agents for any website

Enter a domain, get a voice agent. The platform crawls the site, writes the scenarios and embeds a natural voice assistant with one script tag — unlimited usage, flat price.

talk.yapzek.ai packages the YAPZEK® voice stack as SaaS: scenarios, knowledge bases, tools, MCP servers, analytics, audio logs, audit trails and domain security — managed from one console and billed per domain without token limits.

IntegrationOne script tag
ChannelsWeb voice, chat, phone
PricingFlat per domain
StatusLive
MantisClaw — Autonomous agents from WhatsApp — open source
Open source2026

MantisClaw

Autonomous agents from WhatsApp — open source

A Windows, Linux and macOS agent platform controlled from WhatsApp chats. It researches, writes documents, drives a browser, records meetings and even starts a car — supervised from a group chat.

MantisClaw v1 is published on GitHub. The Go API server bridges WhatsApp, the agent core runs a single tool-call loop with self-healing error recovery, and the execution layer includes an embedded Python kernel, a Playwright browser agent, file outputs and playbooks — all on a local PostgreSQL.

ControlWhatsApp DM and groups
PlatformsWindows · Linux · macOS
DataLocal PostgreSQL
LicenceOpen source
MantisCoder — Repository-aware autonomous coding agent
In production2026

MantisCoder

Repository-aware autonomous coding agent

A desktop coding agent that indexes the whole repository with tree-sitter, plans against the code graph and executes with sub-agents, browser, terminal and git control.

MantisCoder is used internally to maintain YAPZEK®'s own codebases — including NxtEdge — and is offered to enterprise engineering teams that need an agent operating inside their network on large, legacy repositories.

Analysistree-sitter code graph
ExecutionAgentic loop + sub-agents
ControlBrowser · terminal · git
PlatformDesktop, Go + native shell
YAPZEK®

Ready to put AI on your own infrastructure?

Tell us about your environment, your data boundary and your latency target. We will design the pipeline with you and bring it live within days, not months.