Spendly
Developed a predictive expense tracker powered by a custom ensemble forecasting model combining linear regression and exponential smoothing with zero heavy ML library bloat.
> 80%
Forecast Accuracy
Validated on multi-month historical trends
Pure Python
Model Footprint
Zero scikit-learn / TensorFlow overhead
< 2 ms
Inference Latency
Instantaneous client-side chart projections
JWT HMAC-256
Authentication
Stateless, secure session token handling
Operational Constraints & Friction
Users struggle to predict upcoming expenses using traditional manual trackers, leading to reactive budgeting and unforeseen monthly deficits.
Architectural Execution
Developed a predictive expense tracker powered by a custom ensemble forecasting model combining linear regression and exponential smoothing with zero heavy ML library bloat.
Spendly — System Architecture
High-Level Topology & Data Flow
Responsive Web Dashboard & Interactive Canvas Charts
Client requests, sensors & telemetry
FastAPI REST API Layer
ACID Persistence & Read/Write separation
Internal Topology & Data Mechanics
Users record expenditures via a responsive frontend. FastAPI processes entries, calculates seasonal and linear trend weights using custom mathematics, and projects the subsequent month's spending ceiling.
Engineering Obstacles Overcome
Building robust mathematical forecasting models directly in pure Python without introducing heavy, memory-intensive external ML libraries onto low-cost cloud instances.
- ✓Dynamic Expense Categorization & Tracking
- ✓Custom Ensemble Time-Series Prediction
- ✓Interactive Forecast Trajectory Charts
- ✓JWT Stateless Token Authentication
- ✓Lightweight Financial Analytics Dashboard
Planned Enhancements & Roadmap
Project Delivery & Impact
Achieved over 80% forecasting accuracy across historical household financial test sets.