• Led the end-to-end migration advisory for delivery teams transitioning from GitLab and Jenkins to GitHub and GitHub Actions, including target operating model, rollout planning, and technical enablement. 
• Defined and implemented reusable GitHub Actions ?golden path? workflows (build/test/release) aligned to Git Flow, enabling standardized deployments across Maven- and NPM-based services. 
• Designed IaC patterns and landing-zone conventions using Terraform to provision consistent environments and pipelines within an on-prem Cloud Foundry data center and associated platform services. 
• Established security-by-default CI/CD controls by integrating Mend, Fortify, and SonarQube scans into pipeline templates, driving compliance-ready automation and traceable quality gates. 
• Built an AI-assisted support chatbot using Azure AI Foundry to reduce first-level support load, accelerate issue triage, and provide guided answers for common migration and pipeline questions. 
• Implemented ?docs-as-code? governance: maintained operational/runbook documentation in Markdown, automated publishing to Confluence via GitHub Actions, and introduced review workflows for documentation quality. 
• Implemented observability foundations by creating Grafana dashboards for pipeline/platform monitoring and operational KPIs, improving transparency for engineering leadership and customer success stakeholders. 
• Acted as senior integration point between platform, security, and product teams?facilitating workshops, technical decision records, and stakeholder alignment to accelerate adoption and reduce delivery friction. 

• Architected and implemented an Azure Databricks platform to reconcile PDM (Ground Truth) with SRM (Supplier) data, ensuring data integrity for liability and compliance.
• Engineered advanced matching logic utilizing Neo4j, Vector Embeddings, and LLMs to align discordant data structures and identify critical components (semiconductors, rare earths).
• Built automated ELT pipelines (Medallion Architecture) orchestrated via Databricks Asset Bundles for reliable daily data processing.
• Deployed high-quality data products serving both downstream analytics and a custom frontend hosted on Azure App Services.

• Built pipelines to identify and consistently assign customers with the highest potential for purchasing new products and to support product launches.
• Established a Databricks / DevOps / Azure infrastructure for a team of 12 employees.
• Created notebooks in a Medallion architecture to extract, load and transform source data (12 sources).
• Set up orchestration in Azure Data Factory (ADF).
• Implemented a process to send customer datasets to Emarsys (CRM system).
• Introduced monitoring and error handling.
• Deployed an Azure Data Platform (ADF, ADLS Gen3, Functions, DevOps) and Databricks to implement data pipelines, ETL processes and machine learning workflows.

• Engineered a data pipeline to process and filter over 60,000 PubMed abstracts from an initial corpus of 38 million.
• Architected a multi-stage extraction system utilizing LLMs for relationship extraction, specialized NER models, and a custom model for mapping entities to normalized gene identifiers.
• Populated a Neo4j graph database with the extracted entities (genes, proteins) and their relationships to serve as the single source of truth.
• Developed a web-based UI using Databricks Apps, featuring a conversational AI that answers user queries by generating and executing Cypher queries against the knowledge graph in real-time.
• Implemented a ?power-user? mode that exposed the underlying Cypher queries and rendered interactive graph visualizations directly in the UI.

• Authored the official, company-wide AI strategy, defining the long-term vision, governance framework, and strategic roadmap for leveraging AI across the organization in alignment with VIG Group directives.

• Conducted a comprehensive analysis of business processes and led stakeholder workshops to identify, prioritize, and create business cases for high-impact AI initiatives.
• Established a clear framework for the responsible and ethical use of AI, including guidelines on data privacy, model transparency, and risk management to ensure compliant and secure adoption.
• Collaborated closely with senior leadership and board advisors to ensure tight integration and synergy between the newly developed AI and Cloud strategies, creating a unified technology vision.

• Architected and built an end-to-end LLM evaluation framework on the Databricks platform to quantitatively measure the performance of an agent-based extraction system for key contractual data points (e.g., inclusions, exclusions, counterparties).
• Developed interactive Databricks Dashboards and utilized MLflow to track experiment metrics, providing stakeholders with a clear, real-time view of the agent's accuracy and consistency.
• Executed a proof-of-concept for performance improvement by fine-tuning a Hugging Face model using the QLoRA methodology, demonstrating a significant enhancement in extraction accuracy on domain-specific terminology.
• The resulting framework provided the basis for a data-driven approach to iteratively improve the LLM agent, ensuring high-quality, reliable outputs for simplifying contract analysis.

Project 1: 

• Developed a solution proposal and created a detailed project plan.
• Designed a pipeline for data preprocessing and provisioning using Azure AI Search
• AI-assisted optimization of the existing search system by implementing relevance-based hybrid search.
• Integrated generative AI (GPT-4o) with LangGraph to answer user-specific questions based on help documents.
• Conducted training and onboarding for employees to ensure the effective use and maintenance of the system.

Project 2:

• Developed a solution proposal and created a detailed project plan.
• Designed a pipeline for data preprocessing and provisioning using Azure AI Search
• AI-assisted optimization of the existing search system by implementing relevance-based hybrid search.
• Integrated generative AI (GPT-4o) with LangGraph to answer user-specific questions based on help documents.
• Conducted training and onboarding for employees to ensure the effective use and maintenance of the system.

Project 3:

• Designed and built a proof-of-concept data pipeline in Azure Data Factory to process unstructured documents (e.g., publisher data, help articles) following the Medallion (Bronze-Silver-Gold) architecture.
• Orchestrated data transformation workflows by integrating Azure Functions for lightweight processing and Azure Synapse Notebooks for complex, Spark-based transformations of the raw data.
• Implemented the final data loading stage into Azure AI Search, ensuring the data was properly structured and indexed for downstream RAG applications.
• Delivered a comprehensive technical evaluation and cost analysis of the ADF-based solution, which directly informed the strategic decision to leverage the existing Kubernetes cluster with Argo Workflows for the final production implementation.

Project 4:

• Designed and developed deep learning models based on Recurrent Neural Networks (LSTMs, GRUs) to analyze and predict anomalies in multivariate time-series data from server infrastructure.
• Engineered a data pipeline to ingest and process real-time server metrics, including CPU utilization, RAM usage, and network traffic, leveraging Azure Monitor and Grafana for data sourcing and visualization.
• Trained and validated the models, achieving a 90% predictive accuracy in identifying critical failure patterns within a simulated test environment.
• Delivered a comprehensive proof-of-concept that successfully demonstrated technical feasibility and provided key data for a strategic cost-benefit analysis regarding on-premise infrastructure versus cloud migration.

• Evaluated and validated various approaches to automate image retouching, including the development of a concept for neural networks.

• Transferred knowledge through training and documentation to enable project partners to apply and further develop the system.

• Trained and optimized Generative Adversarial Networks (GANs) to implement sub-processes of image editing, such as automated image segmentation.

• Gathered and preprocessed suitable training data in close collaboration with the companies involved in the project.

• Calibrated 3D scanners to ensure a more precise capture of color and depth information.

• Designed and conducted employee training sessions to facilitate the integration of the new technologies into existing workflows.

• Developed, trained, and validated machine learning models (including neural networks, XGBoost, and LightGBM) to enhance the color rendering of 3D scans

• Architected and trained a predictive classification model using boosted decision trees (LightGBM) on a large-scale, highly imbalanced dataset of over 126,000 string theory models.
• Systematically evaluated and benchmarked the performance of various ML algorithms, including Random Forests, SVMs, and Neural Networks, to identify and validate the optimal approach.
• Performed in-depth feature importance analysis to identify the key phenomenological properties used by the model for classification, providing critical insights into the underlying model structure.
• Developed a predictive tool capable of extrapolating from the training data to predict the most probable origin of the MSSM, effectively narrowing a complex search landscape.
• Co-authored and successfully published the complete methodology and findings in the peer-reviewed journal "Progress of Physics," validating the scientific impact of the results.