Dintel

We Build Custom AI Tools & MVPs.

Stop guessing. We turn business logic into functioning AI software. Creators of the Voicyo voice cloning platform.

pipeline.py — RVC Training
import torch
from rvc_modules import VoiceTrainer
# Initialize RVC model configuration
rvc_config = {
"sample_rate": 48000,
"f0_method": "rmvpe",
"hop_length": 512,
"device": "cuda:0",
"epochs": 1000
}
# Initialize training pipeline...
Live Production App

Voice Cloning Marketplace

A fully functional AI voice cloning marketplace built and launched. Features GPU-accelerated RVC training pipelines, real-time inference, and split-payment architecture.

API Inference240ms ⚡
Training Time~10m

Powering our builds

Next.js
Python
Supabase
Stripe
OpenAI
NVIDIA

Proven Execution: Voicyo

We don't just advise. We build scalable platforms with complex revenue logic.

Live Production App

Voice Cloning Marketplace

A fully functional AI voice cloning marketplace built and launched. Features GPU-accelerated RVC training pipelines, real-time inference, and split-payment architecture.

50/50 Revenue Split Logic (Stripe Connect)
Complex marketplace payment architecture
Next.js Frontend + Python/FastAPI Backend
Full-stack production architecture
Dedicated GPU Infrastructure (GCP)
Cloud-based GPU compute on Google Cloud Platform
Training Time: ~10m
Optimized model training pipeline
API Inference
240ms
Fast

Real-time voice inference with sub-second latency

Training Time
~10m

GPU-accelerated model training pipeline on GCP infrastructure

Monetization

Stripe Connect 50/50 Split Logic

Implemented complex revenue sharing between platform and voice creators. Automated payouts, dispute handling, and complete financial reconciliation.

Technical Services

We ship the slide decks and ship code.

MVP Sprint

From idea to deployed app in 4-6 weeks. We skip slide decks and ship code.

Fast Start 4wEarly Demo

Legacy Integration

Injecting LLMs into existing business workflows to automate operations.

API DesignData Pipelines

Audio/Vision AI

Specialized RVC/Computer Vision models with Custom Python/FastAPI Backends for industrial or creative use cases.

Model TuningReal-time Inference

Built by Engineers, Not Salesmen.

Mindaugas Juska

Lead Engineer: Mindaugas

Based in Vilnius, Operating Globally

I bridge the gap between creative signal processing and hard-core software engineering. With a background in professional audio (AudioLab) and full-stack architecture, I build AI tools that don't just work—they perform.

Ready to build something serious?

We only take on 2-3 new clients per quarter to ensure maximum technical attention.