Building and improving databases, acquiring data, ETL/ELT, Big Data pipelines, and deploying cloud services on projects.

Administering infrastructure solutions to improve data models, increase data accessibility, and foster data-driven solutions for clients.

Implementing monitoring solutions to ensure data integrity; working closely with engineers, product managers, and other stakeholders.

Building a Twin Tower architecture-based personalized recommendation system where user and item embeddings are projected in the same vector space; creating the embeddings from a BERT-based model.

Developing image sim algorithms for visual feature mapping for personalized recommendations.

Leading the design and deployment cycle of the entire implementation, achieving a revenue bump of 3% for promoted items.

Created a custom loss function to compare two continuous probability distribution surfaces, enhancing the end-to-end model's efficiency to give better results than Google's Vision OCR for Screenshots by 2%.

Introduced a new vertical for synthetic data generation for machine learning projects.

Led a team of 25 developers to create state-of-the-art ML solutions with a measurable impact on businesses.

Designed a CapsNet-based human and suspicious object detection algorithm for autonomous ATM dynamic surveillance.

Designed and deployed a real-time Twitter hate-speech recognition system based on SVM classifiers with Hyperbolic Tangent Kernels.

Contracted by various US clients to work on ML/DS/NLP projects, joining the top 0.01% of ML freelancers on Upwork.