• Spearheaded research and development efforts for large language modeling (LLM) projects, exploring innovative approaches to improve model performance and efficiency.• Conducted in-depth analysis of model outputs and various parameter efficient fine-tuned methods to optimize model accuracy and generalization.• Data Science lead for various document understanding feature such as Code generation, Custom Key Value, Table extraction and Prebuilt ID model.• Coordinated and led… Show more • Spearheaded research and development efforts for large language modeling (LLM) projects, exploring innovative approaches to improve model performance and efficiency.• Conducted in-depth analysis of model outputs and various parameter efficient fine-tuned methods to optimize model accuracy and generalization.• Data Science lead for various document understanding feature such as Code generation, Custom Key Value, Table extraction and Prebuilt ID model.• Coordinated and led large-scale language modeling initiatives, including LayoutLM-V2 based multi-modal models and Graph Neural Network pre-training.• Made contributions to Oracle's & intellectual property and publication endeavors, including filing 5 patents, with 1 more patent currently in the filing process.• Stayed updated with the latest advancements in large language modeling and incorporated relevant techniques and methodologies into ongoing projects.• Mentored and onboarded new team members, facilitating their understanding of the language modeling architecture. Show less

Implemented a transformer based recommendation module that detects duplicate queries with 92% accuracy. Resulted in 20% of reduction on service team load.

Content personalizing and ranking for VIU users using various Recommender systems from structured and unstructured data combined.

Worked with MIT research student to preparing input feature using PySpark for factors PC-Ratio, Strike, Delta, Moneyness for machine learning model to predict market sentiment in shorter term.Created Data pipeline to support modelling.Performing structured experiment to finding best strategies for targets to use, filling missing values, set of factor to use as input features & approach to use.Tool and model in use: Python, PySpark, Airflow, Logistic regression, Ensemble… Show more Worked with MIT research student to preparing input feature using PySpark for factors PC-Ratio, Strike, Delta, Moneyness for machine learning model to predict market sentiment in shorter term.Created Data pipeline to support modelling.Performing structured experiment to finding best strategies for targets to use, filling missing values, set of factor to use as input features & approach to use.Tool and model in use: Python, PySpark, Airflow, Logistic regression, Ensemble tree(Bagging and Boosting), Forward neural network Show less

As a Technical Specialist-Data analytics my efforts are in the pre-launch & post-launch phase of Philips connected products development.My current responsibility includes analyzing user behavior, providing insights on performance and usage of devices using Adobe Analytics(SiteCatalyst) and apply machine learning and statistics to draw meaningful insights from big data

Worked with Philips Consumer Lifestyle with multiple projects playing various role.