Apr 2026 – Present · 1 mo

Oct 2024 – Mar 2026 · 1 yr 5 mos

• Designed and implemented an optimised cross-region training data transfer, reducing GCS egress cost by 100%.
• Migrated ranker data pipeline to Spark and Databricks, achieving annual cost savings of 3 Cr INR.
• Developed a scalable ETL pipeline using Spark and Flink, increasing model training efficiency.

Jun 2023 – Oct 2024 · 1 yr 4 mos

• Contributed to feature transformation, model inference, and re-ranker layer, ensuring a 99.5% SLA for 180M MAUs through high-throughput, low-latency systems.
• Contributed to a feature-logging pipeline for the MMOE ranker using the BQ API/PubSub, boosting user retention by 0.4% and increasing time spent by 12%.
• Introduced a diversity layer to the final feed, improving content diversity and leading to a 0.1% gain in retention.

Aug 2022 – May 2023 · 9 mos

• Designed and implemented a robust pipeline for the SC AI team to monitor ranker model performance in both online and offline environments across self, random, and all traffic, effectively identifying and addressing bugs in the legacy ranking service.
• Enhanced backend microservices at ShareChat by contributing to logging, feature enhancements, and model serving.
• Implemented inference for the DCN ranker model, resulting in a 0.2% retention gain for 40M daily active users.
• Optimised and consolidated data preparation pipelines, reducing costs by 70%.