My research interests encompass digital signal processing implementation on embedded platforms, ordinary and polynomial matrix algebra algorithm development, and MIMO communication. During my first degree, I implemented an adaptive recursive least-squares filter on a Virtex-5 FPGA for processing functional near-infra spectroscopy (fNIRS) multi-channel signals, enabling functional activity detection. Subsequently, during my M.Sc. study, sponsored by the Commonwealth Commission in the UK, I implemented a physical layer for 5G NR on a Zynq UltraScale+ RFSoC through System Generator. To handle data at the ADC/DAC sampling frequency in the fabric, I utilized parallel processing—known as super sample rate or parallel filtering—enabling the implementation of digital up and down converters circuits using a super sample rate library. This project honed my skills in debugging implementation issues, conducting power and resource usage analysis, and optimizing designs within constrained environments.Sponsored by the Commonwealth Commission in the UK, my Ph.D. work centered on pure signal processing algorithms for addressing broadband sensor array problems. Unlike narrowband signals, broadband sensor array signals exhibit both spatial and temporal correlation. In such cases, the frequency selective channel is represented as a matrix of transfer functions, and the correlation between sensor signals is captured by a space-time covariance matrix which requires polynomial matrix factorization. These factorizations have a wide range of applications including subband coding, speech enhancement, broadband precoding and equalization, beamforming, spatio-spectral MUSIC algorithm, and data compaction.In this project, I expanded existing algorithms by enhancing an analytic eigenvalue extraction algorithm initially designed for 4 sensors to handle 25 sensors. Later, with modifications, the algorithm efficiently computed the eigenvalues of a space-time covariance matrix derived from an array of 200 sensors. Additionally, I improved the extraction of analytic eigenvectors from a polynomial matrix. Initially addressed through non-convex optimization using the dog-leg or Newton method, I introduced an alternating minimization approach for better scalability. Moreover, simplifying the problem into a root-finding task via spectral factorization, I concurrently developed specialized algorithms suited for specific applications requiring partial factorization with the help of Gram-Schmidt, random polynomial projection, and cepstrum-based techniques.