Stefan Lang

- Led end-to-end development of an GenAI-powered conversational platform, boosting user engagement by 25% - Architected and deployed a production-grade conversational AI agent using LangChain and LangGraph for complex, multi-step reasoning, utilized an advanced RAG pipeline with OpenAI models, Tool Calling, Prompt Engineering, MCP servers, A2A agent protocols, serving thousands of daily users. - Built and maintained a scalable FastAPI inference endpoint, containerized with Docker, and automated the CI/CD pipeline using GitHub Actions and GCP, reducing deployment failures by 15%. - Fine-tuned a large-scale speech synthesis model using pretrained feature encoders, resulting in high-quality audio. - Optimized a BARK-based TTS model, achieving a 10x reduction in response time. - Mentored a team of 5 developers, improving code quality and reducing production bugs by 30%.

- Supervised a team developing and integrating NLP models into a customer service chatbot, improving model accuracy by 15%. - Developed a Django based e-commerce application with a React frontend, integrating a multimodal chatbot boosting customer engagement by 25%. - Optimized frontend performance with lazy loading, code splitting, and efficient state management in React, reducing application load times by 40%.

- Developed a novel system for analyzing patient-clinician interactions by integrating FER and SER, achieving 84.5% accuracy on clinical trial data. - Fine-tuned a transformer model for Speech Emotion Recognition from patient interviews, achieving 81.82% accuracy to identify biomarkers for treatment response. - Contributed to an object tracking system using YOLO for analyzing cell morphology in medical imaging, successfully reducing memory usage by 50% while improving real-time detection efficiency. - Enhanced a classification model's accuracy by 20% through advanced visual data preprocessing of diagnostic images, enabling better insights for oncology research.