Parth Modh

• Designed and developed a real-time leaderboard for a season-long fantasy game which is the first and only of its kind in the world.
• This powers any number of simultaneous matches going on for the same fantasy round with a latency of 3 seconds to update the leaderboard and ranks using Spark and Cassandra.
• Built various real-time use cases like player selection percentage; granting real-time rewards, and promotion based on certain criteria; and Observability metrics dashboards for frontend apps relying on click stream events using Flink.
• Developed query parser to parse all the queries run on every OLAP database like Redshift, Athena, and Trino to improve the observability of every data asset.
• Designed an in-house stream processing platform (Streamverse) on top of Flink with Kubernetes as aninfra deployment environment. Here we have integrated various databases and end-points to enrich or push data, which is very useful in building data pipelines or event-based architecture systems.
• Developed topic as a service as part of Streamverse platform with Apache pulsar other than Kafka as messaging queue technology. Worked on POC and load testing Apache pulsar where infra scaling was on focus.

Designed and developed a recommendation engine to serve the best content available as a small team for JOSH and DAILYHUNT.

• Preparing and maintaining user profiling and content profiling of compressed size of 20+ TBs from click stream events using spark and flink data processing frameworks for personalization.
• Developed frameworks to do feature engineering which helps prepare input dataset, train, evaluate, and deploy based on different strategies for ML models like Logistic Regression, Wide and Deep, Frequent Patterns, collaborative filtering, and session-based algorithms on 15 Lakh+ total features of user and content.
• Worked on coding complex inferencing logic of ML algorithms in an optimized way to cater to API requests with minimal latency while having multiple strategies to apply 100+ unique combinations of ML algorithms as part of a recommendation engine.
• Built a system to keep analyzing, monitoring, and authenticating various data to measure the performance of various algorithms and ML models based on AB testing principles.
• Developed real-time, batch, and near real-time pipelines for scoring, trendingness, likes, comments, and other social features for all types of content like videos, news, short videos, and viral images.
• Used spark and flink for data processing and data stores like Kafka, HBase, Redis, HDFS, DynamoDB, Elasticsearch, and Cassandra.

Designed and Built an in-house notification platform which is the primary catalyzer to retain users for any content-based product.

• The platform can trigger notifications in 14 different languages content across Android/IOS/PWA platforms to 100 million+ devices within 30 minutes.
• Developed this platform for operation teams of various Indian languages to pick the best content for notification with multi-layer ACL in place.
• Introduced personalization in notifications for the first time. Designed and developed an automated custom notification flow which has improved engagement 10x. Integrated and enabled various ML algorithms to power personalized notifications for each user.

Contributed to building the most ambitious first-of-its-kind tightly coupled database and application use case on the same layer.

• Combination of RocksDB with integration to use case application layer via RUST. While maintaining high availability, fault tolerance, and minimal latency.