Andrew Nzioki

Software Engineer

I build systems that ship in production.

End-to-end contributor across backend APIs, ingestion pipelines, prediction infrastructure, accountant dashboards, and export workflows.

CV | Resume

Default view is CV.

|
Download CV (PDF)

Andrew Nzioki

Email: [email protected]
GitHub: github.com/Andrew-Nzioki
LinkedIn: linkedin.com/in/andrew-nzioki

Professional Profile

Software engineer focused on financial data systems, AI-assisted accounting automation, and production-grade internal tooling. Demonstrated full-stack and platform-level contribution across backend APIs, prediction pipelines, data ingestion, dashboards, and export workflows in a large collaborative repository.

Technical Competencies

  • Backend Engineering: Python, FastAPI, SQLAlchemy, Alembic, Pydantic, background jobs
  • AI/LLM Systems: Azure OpenAI integration, structured output parsing, prompt design, caching, inference workflow orchestration
  • Financial/Accounting Data: GDPdU parsing, reduced journals, MT940, Plaid, FinAPI, document-to-transaction matching, DATEV-oriented exports
  • Frontend Engineering: React, TypeScript, Vite, SCSS, accountant-facing workflow interfaces
  • Data/Experimentation: Jupyter-based validation and rapid prototyping, seed data pipelines, evaluation workflows
  • Cloud/Platform: Azure Blob integrations, API-driven internal dashboards, operational metrics surfaces

Professional Experience

Software Engineer | Kanzlei21 | Oct 2023 - Jan 2026

Platform and Pipeline Foundations (2023-10 to 2023-12)

  • Helped establish early repository/application structure, runtime scripts, and development scaffolding.
  • Implemented backend pipeline and document-processing modules in the legacy architecture.
  • Added tests and foundational models/schemas as early service capabilities were built.
  • Worked on client identification, receipt/invoice handling, and full-pipeline routing logic.

GDPdU Ingestion and Journal Processing Expansion (2024-02 to 2024-06)

  • Built GDPdU upload and ingestion flows for XML/CSV extraction, column mapping, and JSON persistence.
  • Added/updated schema models and migrations to support GDPdU and journal processing.
  • Integrated process-buchungssatzprotokoll workflows into backend processing.
  • Delivered bank posting matching capabilities and booking prediction generation.
  • Added indexing and historical matching support for reduced journals.

Prediction Intelligence and Integration Breadth (2024-07 to 2024-12)

  • Added two-factor authentication support in backend auth flows.
  • Extended prediction support for Plaid-based financial transactions.
  • Implemented tax code prediction pathways and reliability fixes (including embeddings rate-limit handling).
  • Added prompt caching for posting prediction and shifted to Azure OpenAI structured outputs.
  • Delivered privacy-focused updates such as hiding IBANs in reasoning output.

Production Scale, Split Bookings, Validation, and Exports (2025)

  • Implemented split booking prediction end-to-end across backend and accountant dashboard.
  • Improved split booking generation, prediction stability, and prompt behavior across model transitions.
  • Shipped validation pipeline infrastructure for monitoring prediction quality.
  • Added automation metrics and surfaced validation views in insights dashboards.
  • Implemented DATEV Pro-like export workflows (CSV/ZIP/document flow rules), including production follow-up fixes.
  • Continued production bug resolution in GDPdU parsing, document extraction, export correctness, and UI workflows.

Insights Query Refinement (2026)

  • Updated accounting-firm and automation-related insights queries in dashboard pages to align with evolving data relationships.

Representative Feature Milestones

  • 2024-04-04: GDPdU processing feature.
  • 2024-04-25: Buchungssatzprotokoll integration.
  • 2024-06-04: Booking predictions for bank transactions.
  • 2024-08-29: Plaid prediction support.
  • 2024-11-21: Prompt caching for LLM posting prediction.
  • 2024-11-29: Azure structured output migration.
  • 2025-02-25: Split booking prediction implementation.
  • 2025-03-05: Split booking and model-transition improvements.
  • 2025-05-29: Validation pipeline for prediction performance.
  • 2025-11-15: DATEV Pro-like export workflow.
  • 2026-01-20: Insights query updates.

Working Style Indicators

  • Strong cross-boundary delivery across backend, frontend, data workflows, and experiments.
  • Frequent iteration on production-critical areas (prediction quality, export correctness, ingestion reliability).
  • Consistent use of migrations/models/seed data to support feature evolution and testing.

References

Available upon request.