Backend systems
that survive real scale.

Architecture & implementation

• Architected a business-scoped backend with Django and PostgreSQL, modelling organizations, channel connections, customers, cross-channel identities, conversations and attachments.
• Built a reusable Graphene-Django GraphQL foundation with Relay-style connections, cursor pagination, filtering, sorting, coded errors and field-level permission enforcement.
• Created a provider-neutral AI layer with OpenRouter and Pydantic, including multimodal message schemas, structured tool calls, provider error normalization and model capability configuration.
• Designed per-model token usage, generation time, cost and pricing records so AI interactions can be measured and billed accurately as the platform scales.
• Modelled business knowledge bases, channel-specific customer identities and identity merging to preserve useful context across every customer touchpoint.
• Realtime roadmap: streaming responses through SSE, live conversations over WebSockets with ASGI / Django Channels, Kafka, background processing, and MCP-based tools to communicate with online shops, and external integrations.

Architecture & implementation

• Designed and shipped the platform from scratch with Django + DRF and React + TypeScript, containerized with Docker and served through Nginx + WSGI.
• Engineered a plugin-based payment and exchange layer supporting IRR, USD, TRX and XTR with multiple gateway strategies and accurate cross-currency billing.
• Used Celery workers to synchronize Telegram entities, manage advertising workflows and deliver detailed user notifications outside the request cycle.
• Built a type-safe API layer with Pydantic validation, permission checks and automatic OpenAPI documentation generated from field definitions and docstrings.
• Integrated the Telegram Ads API with real-time pay-per-view tracking, plus AI-assisted compliance, campaign generation, behavior analysis and support.
• Developed an MCP server to expose platform capabilities through a structured interface for AI clients and system integrations.

Architecture & implementation

• Built the product backend with Django and Telethon, connecting account management, ordering, payments and automated Telegram operations.
• Provisioned multi-protocol V2Ray access using VLESS, VMess and Trojan with TLS/XTLS encryption.
• Designed load-balanced, multi-hop infrastructure using HAProxy, Nginx and Apache to maintain service continuity across restricted networks.
• Integrated a native TRX cryptocurrency gateway alongside ten additional international and local payment methods.
• Created a web administration system backed by GraphQL APIs for efficient configuration and CRUD operations.
• Applied distributed routing, WARP and TLS packet-fragmentation techniques to reduce server detection under GFW-style filtering.

Architecture & implementation

• Built the core platform with Django, Django REST Framework and MySQL, supporting multilingual and multicurrency service delivery.
• Used Celery to coordinate long-running order workflows and third-party service processing outside web requests.
• Delivered live order progress through Django Channels, WebSockets and Daphne, including received, remaining and pre-order audience counts.
• Integrated the Instagram and Telegram APIs to automate fulfillment across multiple social-media service types.
• Implemented six local and international payment gateways and used Pillow for platform media-processing needs.