Worked on the apache/arrow-rs repository to improve backend robustness by addressing critical bugs in data-path handling for Parquet metadata decoding and CSV processing. Focused on hardening input parsing and preventing allocator overflows, the work introduced safeguards in the Thrift Compact Protocol Parser and SchemaElement validation, using Rust’s error handling and unit testing capabilities. Added regression and fuzzing tests to ensure stability against malformed or attack-like inputs, and implemented checked arithmetic in the RecordDecoder to prevent panics. All changes were validated with comprehensive test coverage, emphasizing reliability and resilience in backend data processing workflows using Rust and modern testing practices.