During September 2025, Sandish contributed to the pytorch/pytorch repository by developing enhanced signal handling for multiprocessing under SLURM environments. He implemented support for SIGUSR1 and SIGUSR2 signals, making their behavior configurable through environment variables, which addressed process lifecycle challenges common in distributed high-performance computing workloads. Using Python, Sandish focused on robust environment configuration and multiprocessing techniques, ensuring the new feature integrated seamlessly with existing workflows. He also created comprehensive unit tests to validate signal handling across various SLURM scenarios. This work improved the reliability and manageability of PyTorch worker processes, demonstrating thoughtful engineering depth and attention to distributed systems.