- Optimized biometric facial recognition systems by developing a privacy-preserving indexing scheme, improving search efficiency by up to 95% on large non-relational databases.- Developed a decision tree search structure, cutting the computational work of biometric identification transactions by 90%.- Performed machine learning analysis on face recognition systems, providing new insights and visualizations using tools such as T-SNE.- Developed an Autoencoders-based deep learning architecture for facial reconstruction, reducing errors by approximately 1%.- Analyzed and prepared diverse data to enhance biometric data collection and quality filtering processes.- Improved contactless fingerprint recognition systems workload by 50% through the development of deep learning architecture for fixed-length representations.- Implemented a multi-biometric proof-of-concept system that increased speed by 70% and reduced false positives by 10%.- Co-developed a deep learning architecture that ranked 4th in the 2020 mobile sclera segmentationcompetition.- Led stakeholder workshops (IJCB, WIFS, IWBF, BIOSIG) to refine information requirements, resulting in an 80% boost in satisfaction with deliverables.- Performed analysis on GDPR privacy regulations needs, and developed tailored reports, improving decision-making accuracy by 90%.- Used data visualization to deliver actionable insights, cutting insight time by 60%

Vulnerabilities of adversary attacks on facial recognition systems. Research on a solution forthe protection of biometric templates

- Understanding Privacy Protection and GDPR regulations on Biometric Systems.- Implementation of a survey about Privacy Protection.- Transforming stakeholder needs into actionable information

Created data collection pipelines using web scraping technologies to process over 500, 000 data pointsweekly.

Devised abstract syntax tree-based algorithms to automatically interpret ASTM, Kermit and HL7 message protocols from different laboratory instrument vendors, providing robust software solutions in an agile environment.

- Developed supervised classification methods for automatic number plate recognition, enhancing applications for over 10 customers by 40%.- Improved semantic segmentation algorithms by 40% by creating and labelling a semantically annotated dataset for fully connected networks over iris segmentation.- Performed analysis on potential effects of eyewear on iris recognition, revealing a 20% accuracy decline.- Developed new classification methods for detecting agricultural crops in UAV RGB images with up to 96% accuracy.

Worked on Fully Convolutional Networks for Iris semantic segmentation