Backend Engineer (Python)
B2B platform where grain traders, carriers, terminals and financial partners close and execute deals. Lean team, AI-native development.
- Logistics end to end. Deliveries, truck voyages with weighing, a Telegram bot for drivers, vessel voyages and nominations, laytime and demurrage, PDF waybills (TTN).
- Deal execution and terminals. Counter offers with deadlines and extensions, settlement and prepayment, VAT rates for multi-quality contracts, disputed payments, two-sided closing, shipment document drafts shared through chat; order rules such as auto-stopping a sold-out multi-order; terminal appointments with lab checks, quality sampling and Excel registries.
- Data layer. Async SQLAlchemy 2.0 models and more than half of the ~550 Alembic migrations, including a refactor that moved logistics data from deals into a dedicated delivery aggregate.
- Analytics. Warehouse stocks and platform dashboards on PostgreSQL CTE and JSONB queries with multi-currency totals; buy and sell product charts with class and protein filters.
- LLM features. Port documents (NOR, SOF) turned into a deterministic laytime timeline and checked by an eval harness; vessel certificate extraction; domain tools for the LangChain assistant.
- Agentic development. Introduced spec-driven work with Claude Code and BMAD to the team: 80+ specs, tests before code, multi-agent code review.
- Performance and security. Waybill download time cut from ~10 s to 0.2 s; in-house OAuth 2.0 with Google and LinkedIn replacing Auth0; one-time tickets for WebSocket auth.
- DevOps and automation. GitHub Actions deploy over Tailscale, Docker services behind nginx; a grain price pipeline of 10 n8n workflows that scrape marketplaces and Telegram channels, normalize offers with an LLM and NBU rates and feed the platform, with Code nodes kept as JavaScript files under 270+ node tests.
- Python
- FastAPI
- PostgreSQL
- SQLAlchemy 2.0
- Alembic
- Pydantic v2
- OpenAI API
- LangChain
- n8n
- WebSocket
- OAuth 2.0
- Docker
- GitHub Actions
- S3