Mahamad Suhil

Responsible for architecting and development of AI solutions for Financial Markets Operations. Closely collaborated with product owners and key business stakeholders in decision making and designing of AI solutions for different problems. Delivered several OCR, NLP & Computer Vision based AI solutions by training/fine-tuning SOTA deep learning architectures for classification, content extraction and synthesis from (un)structured documents. Managed/mentored a small group of Data Scientists/ML engineers and python developers. Deployed a scalable multithreaded prediction service with queuing and ack/nack. Lead the setup of a complete GPU based production infrastructure (on-premise) from scratch. Developed a generic & re-usable AI-as-a-Service platform for data extraction from unstructured document images/pdfs with all necessary integrations through REST APIs. Regularly delivered tech talks / training programs on the advances of AI.

Developed multiple deep learning based image classification models for classification of ICs & PCB boards. Developed deep learning models for defect region segmentation using autoencoders, modified U-Net and RCNN architectures. Created micro-services in python with queuing and scheduling in multi GPU environment. Explored semi-supervised and unsupervised auto encoder based architectures for image retrieval task. Optimized matching & retrieval time by feature store creation with LMDB & hashing using Maps STL in C++.

Taught core computer science concepts such as Algorithms, Data Structures, Computer Graphics, Programming in C, C++, Operating System Principles to undergrad students. Coordinated students and faculty members in smooth conduction of courses as a Head of the department of computer science.

Design and development of novel machine learning algorithms for unstructured text representation and classification. Published research outcomes in various reputed international journals and conferences such as Elsevier , Springer and IEEE. Mentored/Guided 15+ post-graduate students for their final projects.

Worked on Video Segmentation and shot detection problem. Literature survey and exploration of techniques for video archival and retrieval.

Title: Representation and Classification of Text Data. Developed ML based algorithms for effective representation, feature selection and classification of unstructured text data. Also addressed the problem of class imbalance and detection of unknown classes in a large corpus of text data.

Worked on video segmentation and shot detection problem.