Jul 2022 – Jan 2024 · 1 yr 6 mos

• Key contributor to the Order Assignment Service, the microservice responsible for seamlessly delivering optimized driver assignments for orders across ~1200 cities, operating concurrently with a 30-second interval.
• Reduced Delivery Time breaches in inclement weather by ~2 minutes to improve user experience and order assignment by designing a Realtime Driver Speed signal that responds to environment changes instantly
• Reduced Driver Wait times to improve operational efficiency by implementing time correction features leveraging historical correction factors obtained via Apache Kafka from Python Jobs triggered by scheduled Airflow ETLs
• Enhanced system efficiency by migrating critical API-driven processes to streamlined event-based workflow with MongoDB as primary storage, mitigating the impact of downtimes on external services
• Boosted accuracy by ~5%, improved driver utilization by 0.3 minutes by integrating in-memory Light GBM ML models for distance and time prediction that are fetched on-demand from AWS S3, replacing external gRPC calls
• Enhanced system reliability by eliminating shared DB queries from the service, migrating them to HTTP calls, and implementing Redis fallbacks to handle failures.
• Substantially contributed to the development of Logistics Simulation Service, a robust load testing framework to comprehensively test the entire Logistics Delivery Flow
• Set up the entire data pipeline flow to synchronize and update all relevant SQL and NoSQL databases for simulating and analyzing behavior under various load conditions.
• Implemented mock gRPC servers for essential external services and integrated high-throughput critical App APIs, achieving a peak throughput of ~100k rpm
• Achieved 10x the highest traffic ever recorded in production by enhancing system's capability to replicate production-like traffic peaks

Jul 2022 – Jan 2024 · 1 yr 6 mos

May 2021 – Jul 2021 · 2 mos · Chennai, Tamil Nadu, India

• Achieved an accuracy score of 0.82 by implementing a DL model for semantic segmentation of engine point cloud using PointNet Architecture; augmented the data using multiple orientations to overcome the lack of data
• Automated Finite Element Analysis pre- and post-processing steps by building a multipurpose tool using Python. Delivered a standalone executable file of the tool using PyInstaller to cater to industry requirements.