Skip to main content
Available for commissions & projects

Software Engineer & 3D Engineering Makerdesktop tools3D engineering

High-performance desktop utilities, on-device AI automation workflows, and precision functional 3D engineering prints.

Engineering lab & Maker studio · Dortmund, NRW, Germany
Desktop Apps (Tauri / Rust / C++)3D Design & Printing (Klipper, ASA, PETG, Resin)Local-First AI Automation & Tooling
9
Desktop apps
92+
Offline models
100%
On-device privacy
0€
No cloud fees
Latest News

What's new

Fresh updates, releases and project notes.

Software Showcase

Desktop Apps & AI Tools

High-performance desktop applications built in Rust, C++, and Tauri. From AI video upscaling to intelligent file classification — everything runs locally on your machine.

Active release

Video2X Ultimate Manager

v1.0.0

Enhance video and images locally with AI and GPU acceleration.

Increase image resolution, reduce noise and smooth video motion without sending media to the cloud. The Tauri and Rust application combines AI models, frame interpolation and FFmpeg export. GPU acceleration and a RAM buffer support local processing after the required models are installed.

  • 92+ Offline Neural Models: Real-ESRGAN, Real-CUGAN Pro, Waifu2x CUNet, SRMD-CUDA, RealSR & Anime4K
  • Optical Flow RIFE v4.26: Motion smoothing up to 60/120/240 FPS with CAIN, DAIN & Audio-PTS sync
  • RAM buffer: ImDisk on Windows, RAMDisk on macOS and RAMCache on Linux store temporary frames in memory to reduce SSD writes.
Real-ESRGANVulkan / CUDARIFEFFmpeg
WindowsLinuxmacOS
Perpetual License
29€/ one-time
Zero SubscriptionPay once, own forever
View product details
AI File Sorter
Active release

AI File Sorter

v2.0

Organize photos and documents locally with AI and metadata.

Classify media, find duplicates and organize documents without sending files to the cloud. The PowerShell and Python toolkit combines local vision models, OCR and ExifTool metadata. A SQLite index supports batch processing, while Ollama or LM Studio provides the local AI runtime.

  • Local VLM Vision Analysis: Zero-cloud visual semantic tagging for images and video keyframes via Ollama / LM Studio
  • Perceptual & Hash Deduplication: SQLite index combining dHash/pHash with SHA-256 for instant near-duplicate cleanup
  • Automated Smart Taxonomies: Intelligent hierarchical folder structuring by scene context, event date and camera model
Ollama / LM StudioSQLiteOCRExifTool
WindowsLinuxmacOS
Perpetual License
19€/ one-time
Zero SubscriptionPay once, own forever
View product details
VideoSorterLocal
Active release

VideoSorterLocal

v4.13.2

Organize video collections locally with AI and batch undo.

Categorize videos, add tags and organize folders without sending recordings to the cloud. The Python and PySide6 application combines filenames, metadata, frame OCR and local vision models. BGE-M3 supports content matching, while batch undo lets you reverse file moves after reviewing the results.

  • Multi-stage analysis: Combine filenames, metadata, RapidOCR and local vision models with a file fingerprint cache.
  • Two classification modes: Weight multiple signals or compare sampled frames with BGE-M3 embeddings and voting.
  • Local AI runtime: Use Ollama or LM Studio offline after setup, with automatic VRAM release on exit.
PySide6Ollama / LM StudioBGE-M3RapidOCR
WindowsLinuxmacOS
Perpetual License
9€/ one-time
Zero SubscriptionPay once, own forever
View product details
About Developer & Maker

Engineering Profile & Hardware Lab

AI Software Engineer, desktop systems architect, and hardware maker specializing in high-performance local AI inference, bare-metal tools, and precision 3D engineering.

Open for projects & commissions

Mateusz Wójcik

AI Software Engineer & 3D Maker

Dortmund, NRW, Germany

What I Build & Engineer

I architect and build high-performance native desktop software in Tauri, Rust, and C++, powered by multi-threaded pipelines and direct GPU compute (CUDA, Tensor Cores, Vulkan NCNN). I specialize in 100% offline Local-First AI systems (VLM/LLM vision inference, neural super-resolution, optical flow interpolation), CLI automation suites, and robust PowerShell/Bash workflows.

// Rust// C++// Python// TypeScript// React// Tauri// PowerShell// Bash
Mateusz Wójcik — AI software engineer and 3D maker
Maker & DevDortmund
3D printing3D modellingGraphicsVideo processingLocal AIAutomation

Core Competencies & Tooling

Full-stack mastery across desktop software, AI acceleration, and physical engineering

Native Desktop & Systems Automation

I build native applications with Rust, C++ and Tauri, and automate repetitive tasks with PowerShell and Python. I focus on efficient memory use, responsive file operations and native system integration without unnecessary runtime dependencies.

Neural Video Upscaling & Frame Generation

Deep mastery of FFmpeg 7.x hardware pipelines (NVENC/QSV/AMF), Vulkan NCNN neural engines, Real-ESRGAN/Real-CUGAN super-resolution, and RIFE 4.26 neural optical flow for artifact-free 60/120/240 FPS frame interpolation.

On-Device AI & Vision Models (Local-First)

Production deployment of local Vision-Language Models (VLM), LLMs, and ONNX Runtime / TensorRT engines via LM Studio, Ollama, and custom wrappers. Complete offline inference with zero API subscriptions, zero per-token costs, and 100% data confidentiality.

3D Hardware Lab & Precision Prototyping

End-to-end mechanical prototyping: parametric CAD design in Autodesk Fusion 360, CoreXY kinematics, Klipper/Moonraker firmware tuning, eddy-current bed telemetry, and production manufacturing in technical polymers (ASA, PETG-CF) and 8K UV resin.

Core Principles

  • Uncompromising Privacy (Local-First)

  • Bare-Metal & Native Performance

  • Lifetime Value & Total Ownership

  • Deterministic & Clean Code

Engineering Process

From real-world bottleneck to production-grade release

  1. 01Identify Bottlenecks & Friction Points

  2. 02Architect Native Core & Ergonomic GUI

  3. 03Rigorous Bare-Metal Stress Testing

  4. 04Continuous Refinement & Polished Delivery

Have a custom software project, AI automation challenge, or high-precision 3D engineering commission in mind? Let's discuss your requirements.

Get in Touch
Automation & Dev Scripts

CLI Tools & Automation Workflows

Code examples for local file classification and Klipper integration. Review dependencies and adapt each example before use. Test on file copies; printer commands require supervision and a configuration suited to your hardware.

ai-file-classifier.ps1

AI & Media

PowerShell 7 batch file classifier leveraging local VLM API endpoints for image/video analysis and SQLite deduplication. Privacy-first, fully offline.

PowerShell 7SQLiteLocal VLMFFmpegExifTool
powershell
param(
    [string]$Path = ".",
    [string]$VlmEndpoint = "http://localhost:8080/vision"
)

# SQLite dedup database
$dbPath = "$PSScriptRoot/file_index.db"
$conn = New-Object System.Data.SQLite.SQLiteConnection
$conn.ConnectionString = "Data Source=$dbPath;Version=3;"
$conn.Open()

# Ensure schema
$cmd = $conn.CreateCommand()
$cmd.CommandText = @"
CREATE TABLE IF NOT EXISTS file_index (
    hash TEXT PRIMARY KEY,
    path TEXT,
    tag TEXT,
    category TEXT,
    processed_at TEXT DEFAULT (datetime('now'))
)
"@
$cmd.ExecuteNonQuery()

# Batch process files
$files = Get-ChildItem $Path -Recurse -File |
    Where-Object { $_.Extension -match '\.(jpg|png|mp4|avi)$' }

foreach ($file in $files) {
    $hash = (Get-FileHash $file.FullName).Hash

    # Skip duplicates
    $check = $conn.CreateCommand()
    $check.CommandText = "SELECT 1 FROM file_index WHERE hash = @h"
    $check.Parameters.AddWithValue("@h", $hash) | Out-Null
    $exists = $check.ExecuteScalar()
    if ($exists) { continue }

    # Send to local VLM for classification
    $bytes = [Convert]::ToBase64String(
        [IO.File]::ReadAllBytes($file.FullName)
    )
    $body = @{ image = $bytes } | ConvertTo-Json
    $result = Invoke-RestMethod -Uri $VlmEndpoint -Method Post -Body $body -ContentType "application/json"

    # Auto-organize into semantic folders
    $targetDir = Join-Path $Path $result.category
    if (-not (Test-Path $targetDir)) {
        New-Item -ItemType Directory -Path $targetDir
    }
    Move-Item $file.FullName (Join-Path $targetDir $file.Name)

    # Record in SQLite
    $ins = $conn.CreateCommand()
    $ins.CommandText = @"
INSERT INTO file_index (hash, path, tag, category)
VALUES (@h, @p, @t, @c)
"@
    $ins.Parameters.AddWithValue("@h", $hash) | Out-Null
    $ins.Parameters.AddWithValue("@p", $targetDir) | Out-Null
    $ins.Parameters.AddWithValue("@t", $result.tag) | Out-Null
    $ins.Parameters.AddWithValue("@c", $result.category) | Out-Null
    $ins.ExecuteNonQuery()
}
$conn.Close()

klipper-auto-calibrate.py

3D Printing

Klipper automated bed mesh and nozzle thermal expansion calibration routine. Connects via Moonraker WebSocket, runs probing cycles, and saves optimized profiles.

PythonKlipperMoonrakerWebSocketBed Mesh
python
import asyncio
import json
import websockets

MOONRAKER_WS = "ws://127.0.0.1:7125/websocket"

async def send_gcode(ws, gcode: str):
    """Send a G-code command via Moonraker WebSocket."""
    msg = {
        "jsonrpc": "2.0",
        "method": "printer.gcode.script",
        "params": {"script": gcode},
        "id": None
    }
    await ws.send(json.dumps(msg))
    response = await ws.recv()
    return json.loads(response)

async def auto_calibrate():
    """Run full bed mesh + thermal calibration cycle."""
    async with websockets.connect(MOONRAKER_WS) as ws:
        # Home all axes
        await send_gcode(ws, "G28")

        # Heat nozzle for thermal expansion compensation
        await send_gcode(ws, "M109 S210")
        print("[+] Nozzle at 210C — measuring thermal drift...")

        # Run bed mesh calibration
        await send_gcode(ws, "BED_MESH_CALIBRATE")
        print("[+] Bed mesh calibrated")

        # Measure nozzle offset at temp
        await send_gcode(ws, "PROBE_CALIBRATE")
        print("[+] Probe offset captured at operating temp")

        # Save config profile
        await send_gcode(ws, "SAVE_CONFIG")
        print("[+] Configuration saved to printer.cfg")

        # Park toolhead
        await send_gcode(ws, "G1 X0 Y0 Z20 F3000")

asyncio.run(auto_calibrate())
Tech Stack & Tooling

Tools of the Trade

The languages, frameworks, and tools I use to build everything from low-level desktop apps to 3D printing firmware and AI inference pipelines.

Languages & Frameworks

Rust
C++
Python
TypeScript
React
Tauri
PowerShell
Bash

3D & Engineering

Fusion 360
Blender
OrcaSlicer
Chitubox
Klipper Firmware
CoreXY Tuning
SLA Resin Calibration
ZBrush

AI & Systems

Ollama
LM Studio
ComfyUI
Vulkan NCNN
SQLite
Linux
Apache
FFmpeg
Get in Touch

Custom Engineering & Support

Need a bespoke desktop application, specialized 3D CAD design, or an automation pipeline? Submit your specifications and receive a technical response within 48 hours.

Workshop in Dortmund, NRW: software licenses are delivered digitally. Custom 3D prints can be collected by arrangement or shipped. There is no walk-in shop with public opening hours — write first.

Project Updates & Direct Inquiries

Stay updated on local AI tools, 3D engineering models, and automation scripts — or submit a custom project inquiry.

Confirm you are not a robot