Executive Summary
total delivery time
independent microservices
of production code
CI/CD
full team size
iOS + Android
The Project
A full-stack digital marketplace for the grain commodity market. The platform automates the entire deal lifecycle — from automated price discovery and counterparty verification to contract signing, freight forwarding, and document generation — serving agricultural producers, commodity buyers, and logistics providers.
The system comprises three client applications sharing a common backend: a web portal for sellers and buyers, a freight-forwarder mobile app for field logistics, and a seller mobile app for deal management on the go.
What Was Built
Backend — 15 microservices, each owning a distinct business domain:
- Registration Service — user onboarding, KYC, legal entity verification
- Deal Service — deal lifecycle management, contract generation, status tracking
- Account Service — company profile, requisites, API key management
- Freight Forwarding Service — forwarding workflow, document generation (waybills, expedition reports)
- Best Price Search — basis conversion, freight cost estimation, automated price matching
- Carrier Uberization — ML-based freight cost calculation with seasonal demand modeling; integration with logistics platforms for transport ordering and shipment tracking
- Counterparty Accreditation — document preparation and automated accreditation workflow
- Financial Partner Service — integration with financial organizations for deal monitoring and stage tracking
Services communicate asynchronously via Apache Kafka. Long-running business processes — deal signing, freight forwarding — are orchestrated with Camunda BPMN, enabling reliable execution, error recovery, and process-level observability.
Authentication and role management run on a self-hosted Keycloak instance. Secrets are managed centrally in HashiCorp Vault. Observability is built on OpenSearch with distributed tracing via Spring Cloud Sleuth and real-time alerting to Telegram channels.
Web Portal — Angular 17 / TypeScript. Core features: automated best-price search, deal creation and management, primary document generation.
Freight Forwarder App — Capacitor + Angular + TypeScript. iOS and Android. Three sections: open forwarding requests, active expeditions (with document generation and offline mode with background sync), completed work archive.
Seller App — Capacitor + Angular + TypeScript. iOS and Android. Price dashboard across key buyer bases, one-tap deal execution from mobile, deal status tracking. Targeted at agricultural enterprise managers.
The shared Angular/TypeScript codebase across all three client applications significantly reduced duplication and kept the team aligned on a single component library.
The Method
Same AI orchestration methodology as the Grain Warehouse Registry — scaled to a larger, more complex system.
Three engineers, each acting as an AI orchestrator for their domain. The project combined multiple dimensions of complexity that demanded broader human coverage: a web portal, two cross-platform mobile apps, ML-based freight cost modeling, BPMN process orchestration, and integrations with logistics platforms and financial organizations.
Each engineer owned a set of microservices end-to-end — architecture, implementation, tests, and CI/CD pipeline. AI agents handled code generation, test scaffolding, documentation, and boilerplate; engineers made all design decisions, reviewed all output, and were accountable for every commit.
This project proves the methodology scales. One engineer can orchestrate a focused backend system in days. A small team of orchestrators can deliver a full-scale platform — web, mobile, integrations, ML — in months. The multiplier holds at both levels.
Tech Stack
Comparison with Traditional Dev
| Traditional | This Project (AI-Driven) | |
|---|---|---|
| Timeline | 9–12 months | 3 months |
| Team | 12–18 people | 3 engineers |
| Architecture | Often monolith or ad-hoc | 15 clean microservices |
| Mobile | Separate iOS + Android teams | Single shared Angular/Capacitor codebase |
| Deployment | Manual or basic CI | 15 automated pipelines, blue-green |
| ML features | Separate data team | Included, same team |
Conclusions
15 microservices. A web portal. Two cross-platform mobile apps. ML-driven freight pricing. BPMN-orchestrated business processes. Multiple external integrations. 3 engineers. 3 months.
The complexity here is qualitatively different from a pure backend system. Mobile requires a different discipline. External integrations multiply the failure surface. ML components need their own data and validation cycles. BPMN processes require domain modelling before a single line of code.
AI orchestration handled all of it — at every layer, for every engineer on the team. The methodology doesn't just work for one engineer on a greenfield backend. It works for a team, on a product that spans mobile, web, ML, and complex third-party integrations.
Same principle. Bigger system. Same multiplier.