Focused on distributed systems, event-driven architecture, concurrency, performance, and production reliability.
Exact GPU vector-search engine built from CUDA kernels through C++20, Java JNI, and a Spring Boot API. Includes persistent GPU indexes, FP16 storage, production hardening, and reproducible benchmarks.
Java workshop for distributed transactions, progressing through hybrid logical clocks, MVCC, routing, intents, clock uncertainty, read restarts, and serializable conflict prevention.
Spring Boot decisioning system for policy servicing, claims, premium calculation, risk evaluation, renewals, and auditable business decisions.
I contribute to infrastructure projects where correctness depends on lifecycle, concurrency, RPC behavior, transaction ownership, and failure recovery.
Merged fix preventing the metrics reporter from starting on controller-only nodes.
PR #30973 coordinating query failure with transaction-commit ownership to address an autocommit race.
PR #4263 making cached RPC connection identity endpoint-aware when a server UID moves to a different host or port.
Distributed systems correctness · Concurrency and deterministic race testing · RPC and connection lifecycle · Failure recovery · Performance and reliability · Invariant-preserving regression tests
Java and Spring Boot services with resilient APIs, transaction boundaries, caching, asynchronous workflows, messaging, and production-oriented failure handling.
MVCC, transaction coordination, clock uncertainty, leader/follower behavior, RPC identity, consistency, concurrency, and failure modes.
Working on real bugs in distributed infrastructure instead of limiting engineering work to personal repositories.
Using profiling, benchmarks, SQL analysis, production telemetry, parallel processing, and regression tests to validate performance changes.