Boxwish Media LLP / Spiritzone
Consultant - Data Engineering
2021-12 · Present
Real-time Order Assignment & Dispatch System
PythonFastAPICelery+3 more
- Designed asynchronous real-time order assignment system for dynamic batching and delivery partner allocation
- Handled 30+ orders per minute per instance with real-time agent discovery
- Built Celery + Redis workflows with centralized logging and production-grade reliability
Order ETA Prediction Service
FastAPITensorFlowMLflow+4 more
- Built stateless ML service for real-time delivery ETA prediction
- Designed feature pipelines using historical and real-time signals
- Improved ETA accuracy by 75% using geofence-based caching
Agent-based SKU Onboarding & Validation Pipeline
LangChainFastAPIRedis+3 more
- Built LLM-based agent pipeline for SKU onboarding and validation
- Added observability across validation stages
- Reduced manual effort by 16–20 hours/week
Platform Reliability & Engineering Practices
- Introduced centralized logging and structured error reporting
- Achieved 80–90%+ test coverage using Pytest
- Performed load testing using Locust