We design and build back-office, payment and AI sales systems
Python services, long-running processes on Temporal, single sign-on with Keycloak. We take a project from architecture to production.
What we build
CRM and back office
Sales pipelines, roles and permissions, workflows, and AI agents that move customers through each stage under a manager's control.
Payments and crypto
Crypto acceptance and payouts, key custody, KYC, and architecture for payment aggregators that work with several banks.
Multi-chain crypto payment and custody platformArchitecture for a multi-bank payment aggregator
Integrations without a public API
A stable REST API on top of services that only have a website: real browser sessions, queues and monitoring.
Data and automation
Blockchain indexers, time-series data, and automatic video production from After Effects templates.
TractionEye: net-worth indexer for TON walletsAutomated After Effects video rendering via API
Cases
We don't publish client names. We can walk you through the details, code and demos on a call.
-
CRM, sales automation, operations, 2026
AI sales CRM with LangGraph agents
An omnichannel CRM where an AI agent works each stage of the sales cycle and managers stay in control.
- LangGraph
- Temporal
- Keycloak
-
Fintech, crypto custody, 2023-2026
Multi-chain crypto payment and custody platform
Crypto deposits and payouts on EVM networks, Tron and TON, with keys in an isolated signer and every money movement run as a durable saga.
- Temporal
- Rust
- Keycloak
-
Fintech, payments, 2026
Architecture for a multi-bank payment aggregator
A blueprint the client's team can build from, covering everything from the domain model to the failover runbook.
- Temporal
- Keycloak
- PostgreSQL
-
Integration and automation, 2025-2026
Browser-grade API gateway for a platform with no API
A production REST API and a Kubernetes fleet of browser workers in front of a large consumer web platform that offers no public API.
- Playwright
- Temporal
- Kubernetes
-
Fintech, TON blockchain, 2024
TractionEye: net-worth indexer for TON wallets
An indexer that rebuilds any TON wallet's net-worth history on demand, from native coins to jettons and DEX LP positions.
- TON
- TimescaleDB
- Python
-
Media, video production automation, 2025
Automated After Effects video rendering via API
An API that turns After Effects templates plus data into finished videos on a managed render farm.
- After Effects
- Inngest
- FastAPI
What we build with
- Python
- The backbone of almost every system we build: about 20 FastAPI services, SQLAlchemy and Alembic on PostgreSQL, asyncio, Pydantic, and pytest and Playwright test suites.
- Temporal
- Long-running, money- and state-critical processes in 8+ production systems: crypto deposits and withdrawals, card issuing and KYC, CRM lead screening, AI reply flows and browser-worker fleets.
- Keycloak
- Our standard identity layer in 10+ systems: OIDC + PKCE, token introspection, client credentials between services, branded login themes and custom Java SPIs for PIN login and Telegram Mini App sign-in.
Also in the toolbox: LangGraph and MCP, RabbitMQ, Redis, React and TypeScript, Kubernetes, Docker, OpenTofu, EVM, Tron and TON, Prometheus, Grafana, Sentry.
How we work
-
Diagnostics and demo
We go through the process you want to automate and show a similar solution on demo data.
-
Pilot
We automate one measurable process and check the result together.
-
MVP
We take the system to a first working version with roles, integrations and monitoring.
-
Support
We keep developing the system, watch how it runs and handle incidents.
Questions
What kind of projects do you take?
CRM and back-office systems, payment and crypto systems, integrations with services that have no public API, data processing and automation. Most of them are systems where a mistake costs money.
What stack do you use?
Python on the backend (FastAPI, SQLAlchemy, PostgreSQL), Temporal for long-running processes, Keycloak for sign-in and permissions, React and TypeScript on the frontend, and LangGraph for AI agents.
How does a project start?
With diagnostics: we go through your process and show a similar solution on demo data. Then comes a pilot on one measurable process.
Why don't the cases name the clients?
We don't disclose clients without their consent. We can show the architecture and implementation details on a call.
Tell us what you want to automate
Describe the process in a couple of sentences and we'll suggest a time for a call.