During a two-month period, Kwk236 focused on backend security and CUDA kernel maintainability across calcom/cal.com and pytorch/pytorch. In calcom/cal.com, they engineered targeted SSRF defenses by blocking localhost and loopback addresses, applying explicit blocklists to strengthen internal service protection and ensure auditable, team-reviewed changes. Shifting to pytorch/pytorch, Kwk236 improved CUDA kernel reliability by removing unused shape-check variables from the avg_pool3d_backward path and introducing regression tests to prevent their reintroduction. Their work leveraged TypeScript and CUDA, emphasizing security best practices and robust software testing. The contributions demonstrated thoughtful depth in both security hardening and codebase maintainability for critical systems.