Oct 2023 – Jun 2025 · 1 yr 8 mos · Bengaluru, Karnataka, India

• Helped build an in-house streaming platform using GStreamer and Temporal to support multi-host live streams with low-latency RTMP/HLS delivery
• Worked on re-architecting the event-ingestion pipeline, which reduced the feature drift significantly in the downstream services along with an estimated 30% cost reduction.
• Led the initiative to improve the existing real-time pipeline in the Ads Team by incorporating Flink SQL and Scylla. This project was part of the wider effort to migrate all batch features to real-time.
• Added multiple new features for better model training and inferencing.
• Integrated Monte Carlo + OpsGenie for monitoring/alerting of 250+ features across both Moj and ShareChat.
• Led cost initiatives in the Ads Team and documented the entire Ads pipeline related to event-ingestion, feature generation, and model inference.

Jun 2022 – Oct 2023 · 1 yr 4 mos · Bengaluru, Karnataka, India

• Implemented a real-time event-joiner from scratch to replace the legacy real-time pipeline. This led to 40% cost reduction as we transitioned from Pub/Sub to Redpanda, leveraged Flink for streaming joins using RocksDB and checkpointing, utilized topic partitioning in Redpanda to eliminate Redis, and reduced the throughput of events by 85% using Avro schema and compression, among other optimizations.
• Achieved monthly savings of INR 7M+/month leading cost initiatives in the ShareChat AI team.
• Worked on a pipeline built on Google's ScaNN (approximate nearest neighbor), which expanded the corpus size 10X from 250k to 2M posts and serves 95% of the total platform views.
• Launched multiple user-level A/B tests to test different initialization strategies and leverage user's recent interactions.
• Added ci-checks, grafana monitoring, lightstep (tracing) and profiler to multiple services, which led to 15% reduction in CPU time.
• Mentored an intern on improving a legacy service that powers all the popularity-based CGs in ShareChat.

Jun 2021 – Jun 2022 · 1 yr · Hyderabad, Telangana, India

• - Worked on the CPQ Cloud framework (Configure, Price, and Quote) that provides accurate prices for products helping customers right from tracking the leads to sales to after-sales and support.

May 2020 – Jul 2020 · 2 mos · Hyderabad, Telangana, India

• - Developed a dynamic off-core service called Smart Scheduler, to allow Lightning Scheduler to support multiple scheduling policies including Priority, Load Balancing and Round Robin.