SAHIL

Standard deep convolutional networks provide classification scores without auditable spatial reasoning, preventing clinical auditability and adoption by radiologists.

Constructed a dual-evaluation pipeline producing localized visual heatmaps paired with game-theoretic feature attribution and clinical confidence intervals.

Using native Fetch alone requires developers to implement and maintain recurring networking concerns separately, leading to duplicated logic and inconsistent error handling across applications.

Designed a modular HTTP client that brings essential networking utilities into a unified, typed API, with configurable request behavior and middleware hooks for extensibility.

Crypto payments often require repeated manual action, making recurring transfers and scheduled payouts difficult to manage efficiently. Building automation also introduces challenges around transaction reliability, wallet security, and execution scheduling.

Built a modular payment automation platform that combines wallet connectivity, configurable payment tasks, scheduling logic, and Solana transaction execution in a unified interface.