Hi, I'm
Aryaman

The Challenge: An asynchronous API gateway and caching layer was silently degrading and hanging under sustained load. The codebase was riddled with intentional decoys, such as a synchronous thread pool and a global lock that falsely resembled a classic lock-ordering deadlock.

The Solution: I diagnosed a severe concurrency bug related to partial failure cancellation and dynamic backpressure. When backend requests timed out (CancelledError), the cancellation handler failed to properly notify the asyncio Condition variables, leaving phantom connections checked out. I engineered a robust asynchronous context manager for connection checkout that guaranteed state reconciliation and proper load-shedding during pool starvation.

The Challenge: Debugging a mock distributed system (Gateway, Auth, Database) where all internal clocks were severely skewed. The log data was intentionally chaotic, featuring heterogeneous timestamp formats and idiosyncratic trace_id serializations.

The Solution: I built a custom, highly resilient parsing pipeline to normalize the raw telemetry data. By identifying symmetric RPC round-trips, I mathematically calculated the exact, constant integer millisecond clock offsets for each microservice. I patched a deliberate anomaly where the Database service logged timestamps exactly 1337ms in the future.