Worked on expanding numpy operation support within the keras-team/keras repository’s OpenVINO backend, enabling seamless integration of mathematical and neural network components across backends. Implemented core numpy functions such as gcd using the Euclidean algorithm, with careful attention to type and scalar shape handling, and added support for operations like hanning, histogram, lcm, and kaiser windows. Enhanced backend capabilities by introducing dropout, binomial sampling, and RNN primitives, aligning API consistency and test coverage with other backends. Collaborated across teams to improve code quality and documentation, leveraging Python, numpy, and deep learning frameworks to strengthen numerical workflows and maintainability.