NumPy Experts in Berlin

• Cut first-30-day post-subscription churn 45% to 20% by revamping onboarding and optimizing time-to-value.
• Drove 3x LTV in 6 months through retention and monetization experiments across the customer lifecycle.
• Owned app redesign and feature delivery leading to lifting active-user NPS from 6.3 to 8.5.

• Designed and implemented a hybrid recommendation system (content-based + collaborative filtering)
• Built end-to-end ML pipelines including data processing, feature engineering, model training, and evaluation
• Developed RAG-based LLM systems using LangChain and vector databases for semantic search and knowledge retrieval
• Established MLOps workflows with MLflow for experiment tracking, versioning, and deployment readiness
• Implemented deep learning models (computer vision & classification) using PyTorch and TensorFlow

• Designed and implemented scalable ETL/ELT pipelines using Python, SQL, dbt, AWS and Informatica to ingest data from sources such as APIs, relational databases, and flat files into Snowflake, reducing pipeline runtime by ~30%.
• Migrated high-volume datasets from on-premises Teradata to Snowflake using AWS services (S3, Glue, Step Functions, IAM), ensuring data consistency and integrity.
• Applied Kimball methodology to design star and snowflake schemas, improving query performance and reducing Snowflake compute costs.
• Implemented automated data quality checks using SQL-based dbt tests and the Great Expectations framework to detect anomalies and enforce data correctness before production loads.
• Orchestrated ETL workflows in Airflow using Python and managed code deployments via Git with CI/CD best practices to increase deployment reliability and maintain pipeline uptime.
• Built interactive Power BI dashboards and curated datasets to enable data-driven decision-making for stakeholders.
• Maintained technical documentation in Confluence for ETL workflows, and led knowledge-sharing sessions for new joiners.

• Acted as technical consultant to optimize multi-layer ensemble models combining ResNet, CNN-BiGRU-Attention, and XGBoost.
• Guided implementation of a Logistic Regression meta-learner to solve class imbalance problems, achieving 92.86% accuracy and 0.9644 AUC on PTB-XL and Chapman-Shaoxing datasets.

• Developing methods for monocular 3D facial reconstruction and personalized geometric modelling from mobile imagery
• Building learning-based approaches for facial shape estimation, video-based facial analysis, and privacy-preserving visual learning

• Lecturer for the GenAI Track at the Master School Institute of Technology
• Development of a full-stack AI application (React + Python/FastAPI) for automated supplier product import with intelligent column and category classification (4-layer hierarchical) including human-in-the-loop validation

• Context: Healthcare and laboratory teams required faster document analysis, treatment-planning support, and reliable AI workflows for MR/VR-assisted operations.
• Contribution: Built the AI healthcare platform, model/agent workflows, VR-glasses deployment platform, REST APIs, Next.js/React interfaces, and CI/CD pipelines.
• Impact: Delivered a production-ready AI product foundation that improved clinical document review, supported laboratory automation, and made VR fleet deployment manageable across environments.

Tech: TypeScript, Next.js, Node.js, React, Java, Spring Boot, Python, PyTorch, TensorFlow, Docker, PostgreSQL, OpenAPI, GitLab, GitHub Actions.

• Insurance service provider*

Challenge: Business-critical documents were processed through external OCR providers, with ongoing costs, dependency, and data privacy risks for sensitive insurance data.

Implementation:

• Architecture and production implementation of an on-premise OCR solution with full data ownership
• Methods for recognizing document structures as the basis for automated further processing
• ML-, NLP-, and LLM/VLM-based information extraction, especially from invoices and quotations

Success: Replaced external providers: full data ownership, GDPR-compliant processing, and 75% lower recurring OCR costs per year

Used technologies: Python, Docker, Microservices, FastAPI, PyTorch, Torchvision, MongoDB, MySQL

• Proceeding what was started in 10/2023, offering data science development skills fulltime to international clients

• Helping companies learn more about their existing (unstructured) data, optimize processes and technical systems, and derive solutions for their problems

• Tools and technology used: Python (sklearn, pandas, numpy, Django, sqlAlchemy, pyTorch), Matlab, Docker, AWS EC2, Lambda, S3, SQL, MySQL, Hadoop & Spark, Machine Learning, DNN, AI, Jira, Confluence, Git, CI/CD, GitLab, Jenkins

• Leading accounting team of 6 FTEs
• Training and development of employees
• Automation of accounting processes with SQL, Python and PowerQuery
• Owning year and month end close
• Taxation topics
• Audits
• Netsuite Admin and further development

• Design of an LLM-based agent system to control the ERP software
• Development of agent workflows with LangGraph and PydanticAI
• Planning interfaces between business logic and language models
• Planning agent orchestration
• Prototype development and demonstration

Tools: Python, Pydantic, React, LangChain, LangGraph, Linux

• Project 1: Multi-Tenant SaaS Platform: Data Integration & Pricing Management
• Objective: New development of ELT pipeline (replacement for Java 6 legacy), integration of heterogeneous source systems (CSV, Excel, Email, external DBs), event-sourcing for complete auditability, multi-tenant architecture for tenant-capable data processing
• Challenge: Processing millions of records daily, audit compliance, data isolation between different tenants
• Solution: Stakeholder workshops for requirements analysis, event-driven architecture with Axon Framework and Apache Kafka, AWS services (EC2, S3, Lambda, SQS, API Gateway) for cloud integration, PostgreSQL with tenant-specific schemas for multi-tenant data isolation, REST API design with Spring Boot for external system integrations, comprehensive testing strategy (JUnit, Spring Test, Postman, PACT, ArchUnit)
• Results: ELT performance improved from 30+ min to 1-5 min; 2-3 hours daily saved through workflow automation; 10-20 hours/week saved through event-sourcing auditability; 100% audit compliance; secure multi-tenant data isolation for 10+ tenants
• Project 2: Multi-tenant CRM System Modernization
• Objective: Migration of CRM system (20+ years PHP/MySQL) to Java microservices, Domain-Driven Design implementation, establishment of Test-Driven Development, multi-tenant-capable SaaS architecture for multiple customer tenants
• Challenge: Remodeling complex business logic, no existing test culture, scalable tenant management with data isolation
• Solution: Comprehensive testing strategy (JUnit, Spring Test, Postman, PACT, ArchUnit), multi-tenant architecture with tenant-specific databases, REST API design with Spring Boot for cross-tenant integration, Kubernetes and Docker for container orchestration
• Results: 2× performance improvement; deployment time reduced from 40+ min to 5-7 min; migration without production outages; scalable multi-tenant solution for 15+ customer tenants
• Technologies: Java, Spring Boot 3.x, Angular, Apache Kafka, AWS (EC2, S3, Lambda, SQS, API Gateway), PostgreSQL, Axon Framework, Kubernetes, Docker, GitLab CI, REST API

• Performed qualitative and quantitative analysis of extracted findings, categorizing themes, evaluating methodologies, and assessing study quality and reliability.
• Produced research reports and evidence summaries communicating key trends, gaps, and opportunities to academic advisors or cross-functional teams.
• Presented findings through well-structured visualizations, tables, and narrative summaries to support decision-making and guide future research directions.

• Supported partner centers in adopting AI and LLMs-based predictive models for small-molecule discovery and therapeutic response using their genomic databases, designing failure modes for AI, and reducing feature-engineering time by 25%.
• Performed Python analysis and improved domain motion coverage by 30% through integration of free energy data with conformational modeling, mapping heterogeneous protein states for accurate structure–function analysis.