Digital Execution Strategy Development

• Coached an international team of eight, internal, digital entrepreneurs within the Digital Transformation CoE. Created processes for empowered decision making. 
• Up-skilled team on design thinking methodologies, including design thinking facilitation and qualitative user research methods. 
• Developed digital transformation execution strategy, based on data, people, cloud infrastructure, and budget, alongwith coaching methodologies,  problem statement development, executive buy-in, and sponsorship.

Digital Transformation in Medical Writing

• Performed process mapping, qualitative analysis of internal processes, resulting in business use case descriptions. 
• Used ChatGPT to lead medical writers to do PoC on different uses cases including document generation, redaction, and translation. 
• Performed market scouting, leading to short list of vendors and the development of Request for Proposal (RfP) process to decrease time to market by several weeks.

Lay Document Generation Using RAG System

• Advised Data Scientist and Medical writer on best practices for development of LLM prompts and prompt chaining along with large pdf document ingestion and vector embeddings. 
• This resulted in a RAG system, which enables medical writers to interactively and semi-automatically create lay summaries of clinical trials, based upon the clinical trial protocol.

Analytics Landing Page

• Served as the Product Owner for the analytics landing page, incorporating a recommendation engine and utilizing Large Language Models for enhanced question answering and cognitive search capabilities across dozens of operational dashboards. 
• Created a roadmap for adhoc visualizations of operational data using LLMs, bespoke data models, and OpenAPI specifications.

Clinical Data Catalogue

• Responsible for the development Data Science aspects of a clinical data catalogue, integrating clinicaltrials.gov and in-house data for seamless access and aggregation of clinical trial information. 
• Focused on semantic search capabilities, I developed advanced NLP pipelines and graph data models within a knowledge graph for visualization of unstructured data. 
• My role extended to facilitating the product team?s ownership, addressing technical and business challenges, and aligning with internal stakeholders to ensure data governance compliance.

Interdisciplinary Hackathon Creating a Gamification of NLP Annotation in a Web App

• Led an interdisciplinary hackathon project focused on gamifying NLP annotation for entity relationship models. Responsibilities included desk research, knowledge resource creation for software developers, UX researchers, and external vendors, and guiding data scientists through NLP pipelines. 
• Managed the development of a game that allowed biomedical researchers to validate NLP entity linking models in a gamified approach, enhancing knowledge dissemination in knowledge graphs, gamification, and NLP modeling.

Conceptualization of Bespoke Knowledge Management System Including PoC

• Led the conceptualization of a bespoke knowledge management system, conducting extensive qualitative user research with over 60 company personnel to determine value proposition and feature set. 
• Collaborated with key stakeholders in design sprints to test features, and developed prototypes using open-source NLP algorithms for advanced question answering and topic modeling. 
• Recognized by the board of governors for innovation, and worked with in-house IT for further project development.

Development of Bioinformatics Pipelines for Genome Assembly Algorithms

• Led a cross-functional team of developers and scientists in creating validation and innovative bioinformatics pipelines on AWS cloud services. 
• The project focused on the development of ultra-fast de novo genome assembly algorithms, leveraging cloud computing for enhanced computational efficiency and accuracy. 
• Utilized AWS CDK and orchestrated various AWS services to streamline the bioinformatics pipeline.

Bespoke Drug Discovery and Market Intelligence Web App

• Led the Data Science development of a web application for drug discovery and market intelligence, focusing on complex ETL processes, advanced NLP for entity recognition, and the development of an advanced search feature utilizing biomedical ontologies. 
• Actively participated in user interviews alongside UX researchers, providing expertise on biomedical subject matter to inform user-centric design.

Data Science Hackathon Utilizing Generative AI to Generate Novel Medicinal Molecules

• Led a hackathon project focusing on utilizing generative AI to create novel medicinal molecules. 
• Responsibilities included conducting desk research, creating knowledge resources for non-biomedical data scientists, and leveraging advanced AI models for molecule generation and property estimation. 
• Collaborated closely with the Medicinal Chemistry Department to align efforts and co-create a new digital initiative.

Curation and Reporting of high-dimensional and heterogeneous NGS and Biomedical Imaging Datasets

• Collaborating with the Radiology, Nuclear Medicine, Bioinformatics, and the Oncology departments to collect, aggregate and analyze high-dimensional and heterogeneous datasets stemming from an observational clinical trial in hepatocellular carcinoma patients. 
• I created workflows for curating NGS data sets with multiple online databases. 
• I extracted radiomics features from multi-dimensional imaging data sets and used heatmaps, hierarchical clustering methods, and protein interaction networks to visualizeand report  insights on the project.

Clinical Reporting Strategy of Genetic Variations

• Curating genetic variations of individual patients with online databases and scientific publications. 
• The aggregation algorithms were containerized for interoperability, and were used for creation of genetic variation reports to help oncologists decide about alternative therapy options for patients not responding to the standard of care.

Using Parallel Coordinates to Visualize Hidden Layers in Feed Forward Neural Networks

• Conceptualized an innovative visualisation approach for understanding how feedforward neural networks can be visualized with parallel coordinates, and advised Master's student to develop interactive web app. 
• This approach belongs to the category of explainable AI, and allowed for visualization of hidden layers and separability of classes in hidden layers, leading to an intuitive approach to hyper-parameter tuning.

Development of interactive webapp for visualization of images, no-code creation of neural network architecture, training, and hyper-parameter optimization

• Conceptualized an innovative workflow tool to visualize large imaging data sets, interactively create neural convolutional auto-encoder architectures, train neural networks, visualize latent representation, and perform batch processing for hyper-parameter tuning. 
• I advised a Master's student to develop the corresponding, interactive web app. 
• This is an important tool for no-code development of transfer learning techniques for convolutional neural networks and un-supervised clustering and classification of imaging datasets.

Utilizing Biological Interaction Networks to Aggregate NGS and Molecular Imaging Data, defining highly disrupted pathways

• Conceptualized a technique to map molecular imaging and NGS data sets to KEGG protein-protein interaction networks, and advised a doctoral student to implement the technique. 
• This led to a reduction of the dimensionality of the NGS and imaging features, and allowed for introspection of the indicated molecular pathways, which overlapped in a small cohort of patients. 
• This represents a novel technique for combining disparate biomedical data sets and pathway analysis.