Jayant Sharma

- Led timeseries forecasting and interpolation project with tree-based ML models achieving over 20% improvement compared to mean-based data imputation - Implemented Bayesian hyperparameter optimization workflow, resulting in a 15% reduction in RMSE - Designed and implemented 10+ metrics to evaluate the reliability of time-series forecasting models - Developed advanced methodologies to quantify uncertainty in machine learning predictions, including aleatoric and epistemic uncertainty estimation

- Designed, and deployed deep learning pipelines on images for fabric defect detection saving over 1.01M kg of fabric - Led a project to identify faulty image captures, reducing false positives by 14%, enhancing operational scalability - Accelerated model training by 20% using XLA compilation and implemented model tracking with MLFlow for real-time monitoring

- Knowledge graph construction on historical literature database - Driving research with NLP models like transformers on historical objects - Carried research activities in Database & Information Systems department on database query optimization using RDF triples. - Led experiments with different implementations of a new relational schema called "subject-object bits" in C++.

- Authored article (under review) for IEEE and my Master thesis on drug repurposing and drug adverse event prediction using large-scale electronic health records of more than 4 million patients using SQL and Python - Created an ETL pipeline including data sources from SNOMED, MedDRA, Pubmed, PheWas and other medicine databases on Slurm-managed GPU clusters - Implemented a relational Graph Attention Network (RAGAT) using PyTorch and Tensorflow to perform knowledge graph completion

- Developed graphic models for the Thermal Systems and Buildings division of the Fraunhofer ISE using react.js - Created interactive visualizations of thermal systems using Python’s networkx library, dash framework, and GoJS

- Responsible for credit risk modelling - Implemented classification of loan applications using Logistic Regression and Decision Trees for default risk prediction - Responsible for analysis of loan applicants' profiles and rate of interest • Developed REST-based micro services for equipment insurance and customer history purposes. • Created a feature to enable Business Rules Management in collaboration with Credit Bureaus and Risk Analytics team reducing the mean credit decision time by 35%.

• Wrote new algorithms that increased Auto Decisioning on credit requests by 20%. Won the President's award for implementing this solution. • Collaborated with the Bots team on RPA solutions using the Workfusion tool • Worked to develop new GUIs to better assist Credit Analysts and dealers using our web app • Supported and developed features for John Deere Finance’s critical web application • Served as an SME for a critically essential software component and won the MVP award for the same

- Specialisation in Artificial Intelligence with a final GPA of 1.7 and a potential publication in graph ML. - Master Thesis on large scale Electronic Health Records using Graph Neural Networks. Achieved a grade of 1.1. - Courses on Foundations of Artificial Intelligence, Artificial Intelligence Planning, Machine Learning, Image Processing, Information Retrieval, Bioinformatics, Computational Economics, Mobile Robotics, Deep Learning, Freicar project.

Computer Science(Focused on programming languages and software development)  CGPA - 8.33/10.00 (1.84 in the German grading system)

Non - Medical(Physics, Chemistry, Maths)Final achieved % - 92.00

Physics, Chemistry, Maths, Geography, History and Civics, Environmental EducationFinal achieved % - 91.00 with a scholarship