Ikram

Research Assistant

Uniklinik RWTH Aachen • Temps partiel

Aug 2024 - Dec 2025 • 1 yr 4 mos

Built end-to-end ML systems for clinical decision support in the ICU, working with MIMIC-III/IV critical care data under PhysioNet data-use protocols. - Designed and trained an LSTM-based ICU mortality predictor achieving AUC 0.978 and AP 0.892, with strict patient-level splits, temporal windowing (6–12 h bins), leakage controls, and model calibration. - Built a production RAG pipeline for clinical notes: Bio_ClinicalBERT embeddings, FAISS indexing, DSPy-structured prompt chains, local LLaMA generation; served via FastAPI with an evaluation harness (precision@k, answer faithfulness). - Delivered clinician-facing explainability dashboards using XGBoost and SHAP, improving trust in high-risk mortality flags and reducing black-box concerns. - Implemented model monitoring (prediction drift, feature distribution checks) and automated retraining triggers; containerized all services with Docker and deployed on Azure with CI/CD. - Used DSPy-driven prompt optimization to generate augmented clinical notes grounded in historical data, expanding rare-case coverage in offline evaluations. Stack: Python, PyTorch, scikit-learn, XGBoost, SHAP, LangChain, DSPy, Bio_ClinicalBERT, LLaMA, FAISS, FastAPI, Docker, Azure

Working Student Artificial Intelligence / Machine Learning

aixigo AG • Temps partiel

Jan 2023 - Apr 2024 • 1 yr 3 mos

Shipped LLM-powered features and ML microservices for Aixigo’s Bloxx wealth management platform, working across the full stack from data pipelines to production deployment. - Built the Smart Portfolio Insider : LLM-powered analytics over financial dataframes with RAG, prompt guardrails, streaming I/O, and robustness checks for client use. - Productionized ML microservices (PyTorch, scikit-learn) via Flask/FastAPI and Docker with GitLab/Jenkins CI pipelines; significantly reduced manual deployment overhead. - Authored shell/Python ETL pipelines for large-scale portfolio data processing; improved data freshness for downstream model training and analytics workloads. - Owned automated testing (FitNesse), developer documentation (MkDocs), and Agile delivery (Jira sprints); contributed Java API services for platform integration. Stack: Python, PyTorch, scikit-learn, LangChain, DSPy, Flask, FastAPI, Docker, Jenkins, GitLab CI, Java, SQL

Technical Consultant

Buzzware Tech • Temps plein

Sep 2019 - Oct 2020 • 1 yr 1 mo

Led end-to-end product delivery for multiple client projects, bridging technical execution with business requirements. - Introduced Agile/Kanban workflows across the team, improving delivery cadence and stakeholder visibility. - Designed data warehouse schemas and ETL pipelines feeding Power BI dashboards for client reporting and decision-making. Stack: Python, SQL, Power BI, ETL, Agile/Kanban