Worked on enhancing reliability and performance for the kvcache-ai/sglang repository by developing a configurable timeout feature for the HF3FS client. This feature updated both the constructor and relevant APIs to accept a timeout parameter, ensuring operations fail gracefully if they exceed specified durations. The approach focused on preventing indefinite blocking and improving responsiveness, particularly in multi-threaded backend environments. Addressed a failure scenario caused by lock scheduling on the main thread, which improved fault containment and clarified failure handling under time constraints. Utilized Python and multi-threading techniques to deliver robust backend functionality within a focused one-month development period.