Anirudh Sharma

o Developing a hybrid retrieval pipeline for Industrial Cold Rolling, combining Knowledge Graphs and Retrieval-Augmented Generation (RAG) for context-aware failure analysis. o Built a local solution using Ollama, Langchain, Neo4j Desktop, and ChromaDB (Vector Database) and to overcome the shortcomings of base LLMs in industrial failure management. o Developed a cloud-based solution using Azure AI Foundry, Azure AI Search, and Azure Blob Storage.

o Worked as an Intern in the department of Automation, Industrial IT and Modelling. o Engineered a fully functional Digital Twin of the project “Thermobox” on SIMIT using TIA Portal (PLC Code) and WinCC (HMI). o Replicated real-world behaviour and process values of I/O’s, Alarms, Safety Technology and the Drive technology. o Researched Functional Mock-up Units (FMUs) and analysed the possible implementation use cases in steel production using SIMIT-CTE.

o Worked in the Urban Energy research team, in the EU-Funded project “FLEX4FACT”, developing dynamic building simulations for industrial production sites across Europe. o Developed Python pipelines to extract open source geoinformation, supported by CI/CD workflows using GitHub Actions. o Conducted spatial analysis using QGIS to identify industrial optimizations. o Automated the building polygon generation using the OpenStreetMap API.

o Built an end-to-end ML pipeline to predict AUV thruster failures by detecting early degradation patterns in sensor telemetry. o Trained a Random Forest classifier on sensor data to identify pre-failure sensor signatures, achieving 81.3% weighted F1-score on test data. o Containerized the pipeline using Docker and automated deployment via GitHub Actions CI/CD. o Deployed the model to Azure App Service with a minimal FastAPI interface for real-time inference.