“Real-time data, observability, and the firehose of IPL traffic.”

• Built Dream11's Winners Page using Apache Spark and Flink for real-time data aggregation across millions of users, handling concurrent contest settlements during peak IPL traffic.
• Contributed upstream OpenTelemetry Java instrumentation for Vert.x 3.9, MySQL, Redis, and AWS SDK — patches merged into the official OTel repo, enabling automatic trace propagation.
• Deployed and tuned a SigNoz + ClickHouse observability stack handling ~1 GB/sec ingestion; extended the internal fork with delta temporality, histogram quantiles, and custom aggregations.

• java
• vert.x
• opentelemetry
• clickhouse
• kafka
• kubernetes

“Built the rewards platform powering 75% of India's UPI volume.”

• Re-architected the rewards engine into a config-driven platform supporting milestone, randomizer, fixed, and referral strategies, enabling zero-dev launches across UPI, BBPS, and investment products.
• Certified the system at 6,000 RPS — covering 75% of India's UPI volume — by resolving thread contention and DB latency bottlenecks via JVM profiling and query optimization.
• Cut reward-eligibility latency from 800 ms to under 120 ms by optimizing critical DB queries and introducing Redis caching with appropriate TTLs and invalidation strategies.

• java
• spring boot
• spring webflux
• redis
• postgres
• kafka

“Real-time payments pipelines and an 8× faster merchant analytics API.”

• Built a distributed real-time transaction data pipeline consuming events from multiple sources and fanning out to analytics, full-text search, and the merchant-facing portal.
• Reduced merchant analytics API response time from 4 s to 0.5 s (8× improvement) by integrating Apache Druid, enabling sub-second OLAP queries on high-cardinality data.
• Designed Kafka consumer groups, partitioning strategies, and dead-letter queues for at-least-once delivery with idempotent consumers — critical for a payments system where data loss is unacceptable.

• java
• spring boot
• apache druid
• elasticsearch
• kafka
• aws s3

“Meta-programmed a GraphQL QA tool that replaced a manual, error-prone process.”

• Built an internal tool using meta-programming to automate GraphQL query generation, execution, and response validation — replacing a manual, error-prone QA process for schema changes.
• Designed the tool to introspect the GraphQL schema at runtime and auto-generate test queries for new types and resolvers, broadening coverage without hand-written tests.
• Implemented response validation comparing actual API responses against expected shapes derived from the schema, catching contract regressions before they reached production.

• go
• graphql
• meta programming
• api testing

“Step into the position the moment before chess history was made.”

Replay Chess turns the most-talked-about positions in chess — from Agadmator's catalogue, classic games, and instructive puzzles — into playable boards you can step through, share with friends, and analyse move-by-move against the engine. It's the fastest path from “how would I have played here?” to actually finding out.

• 6,000+ pause-the-video positions from Agadmator's channel, ready to play out.
• One-click shareable links — challenge friends to find the move you found.
• Engine analysis comparing your move to the optimal line, every ply.
• Browse iconic moments from chess history and pick up where the masters left off.