AI and SaaS Product Development: Voice Agents, LLM Automation, Backends

SaaS products and AI features built by one team: Modern Voice in production in two months, AI agents in a CRM, email triage and test automation.

AI and software products, built and run

We build SaaS products and put AI into them. Modern Voice, a voice AI sales trainer, went from nothing to production in two months with three engineers. For other clients we added AI agents to a CRM, built email triage linked to warehouse stock, and wrote computer-vision test automation. The people who design the product also write the code.

Who we build for

  • Founders with a product idea who need a working version fast.
  • B2B SaaS teams that want AI features in a product they already run.
  • Small companies that want AI to handle part of daily work: email, replies, CRM updates.

Work we have shipped

Modern Voice: voice AI sales training for insurance agents

Agents practise calls with AI buyers that answer by voice, then get a score against the script. We built the architecture, the backend, the web app and the AWS infrastructure. A Spring Boot service bridges OpenAI Realtime and ElevenLabs over WebSocket, and Stripe handles subscriptions with reconciliation of payment events. The product launched on 1 July 2025 and has 100+ agencies and 400+ users.

Welcome Entertainment: AI agents in daily work

AI agents in the CRM, in replies to customer requests and in the company’s own product work. The client estimates product creation became 8–12 times faster. That figure is theirs, not ours.

UGOSKATE: email triage and drafted replies

We have run the UGOSKATE website and CRM since 2023. We added n8n automation and AI that sorts incoming email, drafts replies and checks warehouse stock.

TestUp: computer-vision test automation

From 2021 to 2025 we worked on a no-code testing tool for a German SaaS company. It tests applications through the screen, using OpenCV, Kotlin and Spring, with C++ moved to Rust.

Tuc-Tuc: a secure company messenger

An encrypted messenger for small and medium companies, built on Matrix, with calls on LiveKit. One SvelteKit and Capacitor codebase serves iOS, Android and the web. The app is in the App Store and on Google Play.

Parcel Manager: courier app

An Android app for couriers handling 50–400 parcels a day, with master packs, nested parcels and offline sync.

What we build

  • Voice and LLM agents with a handover to a person. See AI integrations.
  • Product backends in Java and Kotlin on Spring Boot, with Stripe billing.
  • Web and mobile apps: Angular, SvelteKit, Kotlin Multiplatform, SwiftUI.
  • Automation in n8n, linked to your CRM, email and warehouse data.
  • AWS infrastructure for the product.

Stack, next to where we used it

  • OpenAI Realtime + ElevenLabs + Spring Boot WebSocket bridge: Modern Voice.
  • Stripe subscriptions with reconciliation: Modern Voice.
  • n8n + AI email triage: UGOSKATE.
  • OpenCV + Kotlin + Rust: TestUp.
  • Matrix + LiveKit + SvelteKit + Capacitor: Tuc-Tuc.

Who works on this

  • Petro Gordiievych (founder and architect). Designs the architecture of most of our products; 15+ years in Java and Kotlin; ex-Wise.
  • Svitlana Spinatiieva (tech lead). Java and Kotlin, Spring, Kafka, Angular.
  • Ivan Butakov (senior backend engineer). High-load services in Java, Kotlin and Rust.
  • Vasyl Lukinchuk (senior full-stack engineer). Backends, Angular and React, Kotlin Multiplatform.

Modern Voice was built by two developers and a tech lead.

Where this approach does not fit

Realtime voice from hosted providers sends audio to those providers. If your rules forbid that, the design changes: other models, other hosting, a different cost. A customer-facing agent also needs a handover to a person and strict limits on what it may say.

How to start

Start with an AI prototype sprint: one engineer, one week, €2,000, a working version on your data. If it works, the next step is a fixed-scope MVP or a dedicated squad.

Questions we get

Can we start with something small?

Yes. The sprint exists for that: one week, a fixed price and a working result you can show.

Who owns the code?

You do, by contract.

Can you add AI to a product we already run?

Yes. TestUp and UGOSKATE are existing products and systems that we extended; we begin with a review of what is there.

How to start

Three ways in, each with a fixed scope and a fixed price. Pick the one that matches how far along you are.

Free 30-minute technical call

No cost

Talk your idea, current setup or technical problem through with an engineer. You leave with a concrete next step, whether or not you work with us.

Book a technical consultation

AI prototype sprint

€2,000

One engineer, one week, a working MVP you can use in your business.

Start a sprint

Dedicated squad

from €10,000/month

2–3 senior engineers on your product, European hours, 50 % upfront.

Discuss a squad

Let's discuss your connected product

Tell us what you are building, connecting or modernizing. Our Core Team will review your idea, current setup or technical challenge and suggest the best next step.

The Core Team is the group of senior engineers you talk to first. When a project needs more people, we bring in other teams and specialists we have worked with before, and we tell you who they are before work starts.

Choose how to start

Book a technical consultation

Schedule a call with our engineering team to review your idea, current setup or technical challenge.

Book a technical consultation

Send project details

Tell us about your product, connection or modernization task and we will suggest the best next step.

Send project details