FastAPI Construction FinTech Microservices

Project Description

I’m building four independent FastAPI services, each wrapped in its own Docker image and invoked exclusively through server-side Next.js proxy routes (no direct browser calls, no cross-service authentication, tenant_id filtering on every query). The services and their duties are already defined; I need them implemented, containerised, and ready to run anywhere Docker runs.

Service M1 – GAEB Pricing
• Parse GAEB XML with lxml/XPath.
• Enrich German construction text and launch async LightGBM jobs to price items.
• Produce Excel workbooks, printable PDF offers, and GAEB X84 files.
Target: reduce tender pricing from two days to twenty minutes.

Service M5 – Model Hub
• Host nine live models (margin, CPI-based risk alert < 0.90, waste, quality, win probability, price forecasts, cash-flow P10/50/90, anomaly detection, Bayesian margin refresh).
• Provide weekly automated retraining with validation and automatic rollback on degraded metrics.

Service M6 – DIN Calculations
• Deterministic DIN math for twelve trades (areas, waste factors, skirting, etc.).
• Bin-packing optimisation and REB export for quantity take-off precision.

Service M9 – Financials & Compliance
• Earned-value metrics, PDF invoices compliant with §14 UStG, DATEV exports, and fully populated ZUGFeRD/XRechnung e-invoices.

Common build requirements
• FastAPI 0.110+ with pydantic-v2 models.
• Async SQLAlchemy on PostgreSQL, Redis for task queues, Prometheus metrics, and OpenTelemetry tracing.
• Each service ships as a self-contained Docker image with health, readiness, and /docs endpoints enabled.
• CI pipeline must run unit tests, type checks (mypy), and black formatting before pushing a version-tagged image.

Acceptance criteria
1. Calling the dedicated route of each container returns the correct output format (Excel or XRechnung, depending on the service).
2. Weekly retraining pipeline in M5 can be triggered manually and rolls back automatically if validation AUC drops > 2 %.
3. All containers pass docker-scan with no critical vulnerabilities.
4. End-to-end test through the Next.js proxy validates tenant_id isolation.

If you’re comfortable delivering production-grade FastAPI in Docker—and like the mix of GAEB, ML, DIN maths, and German e-invoicing standards—I’d love to hear how you would approach the build and timeline. Show More

Attachments

Freelancers Bidding (0)

  • This project has no proposals yet.
    Be the first to place a bid on this project!