Bengaluru Streams

Thanks to RisingWave for sponsoring the F&B for the meetup!

🕒 Schedule :

10:00 am - 10:20 am: Welcome & registrations
10:30 am - 11:15 am: `Real-Time Analytics at Scale using RisingWave and StarRocks` by Sri Charan Sirpa, Tech Lead at KaptureCX
11:20 pm - 12:00 pm: `Scaling Backend Systems at PhysicsWallah: From Pipelines to ClickHouse and Kafka` by V Santhosh Kumar, SDE-2 at PhysicsWallah
12:00 pm - 12:15 pm: Networking break
12:15 pm - 1:00 pm: `Architecting enterprise database synchronisation using CDC and Kafka Connect` by Balaji K, Lead Platform Engineer at Platformatory Labs
1:00 pm - 2:00 pm: Lunch & Networking

🎙️Talks:
Real-Time Analytics at Scale using RisingWave and StarRocks
Speaker: Sri Charan Sirpa, Tech Lead – Data Platform Team at KaptureCX
About the talk: Building real-time analytics is no longer just about fast dashboards—it’s about processing streaming data continuously and serving low-latency analytical queries at scale.
In this talk, I’ll walk through how we design real-time analytics using RisingWave for stream processing and StarRocks as the serving OLAP database. I’ll explain how streaming data flows from event sources into RisingWave, how materialized views are created and maintained in real time, and how StarRocks enables fast analytical queries across multiple business use cases such as Ticketing, VoiceBots, and QA systems.
The session focuses on practical architecture choices, data modeling strategies, and performance considerations when running real-time analytics in production.

Scaling Backend Systems at PhysicsWallah: From Pipelines to ClickHouse and Kafka
Speaker: V Santhosh Kumar, SDE-2 at PhysicsWallah
About the talk: In the session, I’ll walk through how our video engagement system at PhysicsWallah evolved from a simple, state-based model to a high-throughput, near real-time event pipeline handling hundreds of millions of events per day.
We’ll explore why our early assumption — that lecture completion was a sufficient proxy for engagement — worked at small scale but broke down as we grew. As student behaviors diversified, our binary progress model flattened meaningful learning patterns into identical signals, exposing deeper limitations in how we captured and processed data.
I’ll cover the architectural transition from state tracking to event-driven design, the challenges we faced while scaling Kafka consumers, the impact of silent data loss and rebalances, and why we eventually rethought our execution model to prioritize predictability under sustained load.
Beyond the technical journey, this talk will focus on key engineering lessons around signal integrity, system design under pressure, and why preserving behavioral truth becomes a product-critical responsibility at scale.

Architecting enterprise database synchronisation using CDC and Kafka Connect
Speaker: Balaji K, Lead Platform Engineer at Platformatory Labs
About the talk: In this session, I will present a real-time data synchronization use case implemented to replicate data between two Oracle databases. The pipeline uses the Oracle CDC Source Connector to capture database changes from the source Oracle database and publish them to Kafka topics, and the JDBC Sink Connector to write the data into the destination Oracle database. While the overall architecture works well for standard relational columns, handling LOB data types such as NCLOB introduced additional complexity. The Oracle CDC connector publishes LOB data in byte format within Kafka topics, which requires careful handling of character encoding and data reconstruction before writing it back to the destination database.
In this talk, I will briefly walk through the architecture, highlight the challenges encountered while processing LOB columns, and explain the approach used to reconstruct the data using custom Kafka Connect transformations.