Pooja Deshmukh

•Investigated non-deterministic failuresin a multi-threaded Linux-based system, tracing race conditions acrossshared state and restoring stable

execution across workloads exceeding 1k concurrent tasks.

•Re-architected a distributed academic execution framework, introducing synchronization primitives that reduced execution inconsistencies and eliminated intermittent failures observed during stress testing.

•Leveraged LLM-assisted debugging workflows toaccelerate root causeanalysis of complexconcurrency bugs, reducing investigation time by 30% and improving reproducibility of edge-case failures.

•Owned a high-impact production incident in a distributed service handling 20k requests/min, isolating thread contention and reducing latency

spikes by 42% underpeak load.

•Engineered a debugging framework for multi-threading and deadlock scenarios, cutting mean time to resolution from hours to under 30 minutes during critical outages.

•Reworked inefficient SQL query execution paths in PostgreSQL, improving response times for complex joins by 35% and stabilizing performance under peak transactional load.

•Hardened backend services on Linux, introducing failure recovery strategies that improved system uptime from 97.8% to 99.5% during cascading failures, including strengthening Database Replication reliability.

•Resolved severe thread pool exhaustion ina high-throughput service( 15k concurrent users), redesigning concurrency handling and restoring

stable throughput during traffic surges.

•Led root cause analysis for recurring production outages caused by MySQL transaction deadlocks, restructuring isolation strategies and reducing failure frequency by over 55%.

•Optimized core Java-based distributed components, redesigning data structures and memory usage patterns to improve processing efficiency by 28% under sustained load.

•Spearheaded cross-service debugging during outages, correlating logs across distributed nodes to restore data consistency guarantees under partial system failures.