May 2023 – Aug 2024 · 1 yr 3 mos · Gurugram, Haryana, India · On-site

• ◦ Worked on two highly-scaled, order-critical services: Locations and User, managing and propagating location and user data to multiple downstreams and app, with peak throughputs of 1M and 5M RPM respectively.
• ◦ Built a reverse geocoding pipeline for location queries based on lat-long pairs, supporting current and nearby use cases. This involved ingesting in-house entities like POIs, Zomato places, and Localities on DynamoDB by breaking the polygons into S2 cells for spatial indexing. This helped run various location-specific campaigns on app and improved address text detection(reducing rider drop location errors to 12%).
• ◦ Migrated the core locations API responsible for location autodetection on homepage, add/update address flows, and map pin changes on app from monolith to location service, reducing its response time from 350ms to 130ms. Added support for address pinning, prompts for location GPS issues, location-specific address templates and background address correction, further reducing rider drop location errors to 7%.
• ◦ Developed address flow for train ordering using railway station POIs for flow identification, address text detection, and displaying enroute stations list for preordering. Expanded POI coverage for railway tracks, which, combined with in-app chaser notifications, led to Rs1 Crore monthly increase in train order revenue.
• ◦ Added a dietary preference toggle in the app updating profile constructs on user service to help display veg-only restaurants and dishes to users. Built a location specific scaling mechanism of this feature.
• ◦ Built load shedding mechanisms and migrated sync gRPC ingestion flows to Kafka events for reliability.

Jun 2022 – May 2023 · 11 mos · Gurugram, Haryana, India · On-site

• ◦ Single-handedly led the foundation of Weather Microservice in Go from scratch to mitigate revenue loss during extreme weather conditions(such as rains and high temp.), improving overall UX. Integrated essential tools like Kafka, SQS, Redis, Localcache, HTTP Egress server, GRPC server, MySQL, Newrelic, Grafana and enhanced maintainability with automated dependency injection pattern within service. Migrated critical weather flows from monolithic codebase to this service post successful shadow mode operation.
• ◦ Built the pipeline for collecting and ingesting weather data from AWS weather stations, external APIs(Climacel), incentivized rider video uploads, and agent QC inputs. Developed rain mode’s trigger,
• continuation, and propagation mechanisms to multiple downstreams. Further refactored binary rain signal flags to intensity X ageing (rain accumulation) buckets using Apache’s Flink-based aggregation.
• ◦ All these efforts improved rain detection accuracy by 60%, enhancing rain-related stress handling on restaurant throttling, customer/rider surge control, and rain app banners, boosting RMS to 2%, NM by Rs5/rain order, 33% lower down surge payout with Rs20 Crore savings in 2023 monsoons. Helped open-sourcing AWS weather station data at Zomato’s https://www.weatherunion.com/, India’s first crowd-sourced weather data platform.
• ◦ Enhanced the weather pipeline by integrating temperature data for heat metrics collected from AWS weather stations and introduced prediction context for rains and heat on a mealslot basis enabling proactive rider communications and revenue impacting surge value configurations on the (t-1)th day.