← Back Creator & Architect · Backend Engineer

Aevum

A backend-centric personal-finance platform with a real-time financial accountability engine — track spending, budget, forecast recurring bills, and turn everyday discipline into automatic savings.

What it does

Everyday discipline, into savings

  • Track spending and manage per-category budgets.
  • Forecast recurring obligations automatically from transaction history.
  • Import and auto-categorize bank / UPI statements.
  • A small self-imposed consumption tax on qualifying spend is set aside into a dedicated savings account; budget breaches add a marginal penalty on top.
Under the hood

The engines

Financial Accountability Engine

Real-time weekly bill computation, a write-once tax ledger, a five-state bill lifecycle, breach penalties, and frozen-period adjustment posting.

Recurring Intelligence Engine

Detects repeating obligations from history, infers cadence, forecasts upcoming bills without writing transactions, and reconciles against reality.

Statement Processing & Categorization

Async ingestion → normalization → auto-tagging with dependency-safe downstream refreshes that keep budgets, analytics, and forecasts consistent.

Auth & Security

JWT auth, device-aware sessions, new-device verification, TOTP 2FA, recovery flows, and administrative safeguards.

By the numbers

A production-scale build

18
feature modules
41
data models
112
REST endpoints
699
backend tests
6
scheduled workers
A 224/225
maintainability (radon)

Backend hot paths, measured on a Postgres testcontainer (indicative, not contractual): single-transaction tax recalc ~51 ms · transaction list (25/page) ~3.8 ms · categorize 500 statement rows ~22 ms. Benchmark suite: 69/69 modules green.

My contribution

Designed & built solo

I owned Aevum end-to-end — architecture, financial engines, security, background processing, testing, and documentation.

  • Architected Aevum as a service-oriented modular monolith with strict domain boundaries and cross-module access through public service contracts.
  • Built the taxation engine that recalculates weekly obligations on every mutation, maintaining a write-once ledger with adjustment-based corrections.
  • Designed an inference engine that forecasts recurring bills without writing transactions, then reconciles forecasts against reality.
  • Kept correctness under statefulness — pure-engine/persistence separation, write-once ledgers, and frozen-period adjustments so history is never silently mutated.
Stack

Built with

  • Python
  • FastAPI
  • PostgreSQL
  • SQLAlchemy (async)
  • Alembic
  • Redis
  • APScheduler
  • OAuth
  • JWT
  • TOTP 2FA
  • Pydantic
  • Docker
  • Pytest